Time-frequency analysis reveals an association between the specific nuclear magnetic resonance (NMR) signal properties of serum samples and arteriosclerotic lesion progression in a diabetes mouse model

Kanako Yui; Yoshimasa Kanawaku; Akio Morita; Keiko Hirakawa; Fanlai Cui

doi:10.1371/journal.pone.0299641

. 2024 Mar 8;19(3):e0299641. doi: 10.1371/journal.pone.0299641

Time-frequency analysis reveals an association between the specific nuclear magnetic resonance (NMR) signal properties of serum samples and arteriosclerotic lesion progression in a diabetes mouse model

Kanako Yui ¹, Yoshimasa Kanawaku ^2,^*, Akio Morita ³, Keiko Hirakawa ⁴, Fanlai Cui ²

Editor: Jorddy Neves Cruz⁵

PMCID: PMC10923453 PMID: 38457384

Abstract

Diabetes causes arteriosclerosis, primarily due to persistent hyperglycemia, subsequently leading to various cardiovascular events. No method has been established for directly detecting and evaluating arteriosclerotic lesions from blood samples of diabetic patients, as the mechanism of arteriosclerotic lesion formation, which involves complex molecular biological processes, has not been elucidated. “NMR modal analysis” is a technology that enables visualization of specific nuclear magnetic resonance (NMR) signal properties of blood samples. We hypothesized that this technique could be used to identify changes in blood status associated with the progression of arteriosclerotic lesions in the context of diabetes. The study aimed to assess the possibility of early detection and evaluation of arteriosclerotic lesions by NMR modal analysis of serum samples from diabetes model mice. Diabetes model mice (BKS.Cg db/db) were bred in a clean room and fed a normal diet. Blood samples were collected and centrifuged. Carotid arteries were collected for histological examination by hematoxylin and eosin staining on weeks 10, 14, 18, 22, and 26. The serum was separated and subjected to NMR modal analysis and biochemical examination. Mice typically show hyperglycemia at an early stage (8 weeks old), and pathological findings of a previous study showed that more than half of mice had atheromatous plaques at 18 weeks old, and severe arteriosclerotic lesions were observed in almost all mice after 22 weeks. Partial least squares regression analysis was performed, which showed that the mice were clearly classified into two groups with positive and negative score values within 18 weeks of age. The findings of this study revealed that NMR modal properties of serum are associated with arteriosclerotic lesions. Thus, it may be worth exploring the possibility that the risk of cardiovascular events in diabetic patients could be assessed using serum samples.

Introduction

Background

Diabetes can cause arteriosclerosis, primarily due to persistent hyperglycemia, and arteriosclerosis is associated with subsequent development of various cardiovascular diseases [1, 2]. The detection of arteriosclerotic lesions at an early stage, before progression to a pathological condition with a high risk of cardiovascular events, is important in diabetic patient management. To diagnose a disease or pathological condition, it is important to determine the specific properties of the disease or pathological condition from clinical specimens such as serum. In a clinical setting, other tests (e.g., imaging [3–5] or functional tests [6, 7]) are needed in addition to blood tests to detect arteriosclerotic lesions. However, in Japan, there are no clear criteria for the screening of arteriosclerotic lesions, even though comprehensive risk management is recommended in clinical practice guidelines [8, 9].

There are many reports [10–14] on the changes in specific components of blood associated with arteriosclerosis and cardiovascular events in patients with diabetes mellitus. The ability to evaluate arteriosclerotic lesions using blood alone could make the management of arteriosclerosis more practical in clinical settings. Unfortunately, no tests have been established that can directly detect and evaluate arteriosclerotic lesions solely from blood samples of patients with diabetes mellitus, as the detailed mechanism of arteriosclerotic lesion formation, which involves complex molecular biological processes, has not been elucidated [15–17].

Serum contains many hydrogen atoms that make up water and other molecules. The interactions within and between molecules caused by hydrogen atoms affect the structure and function of molecules contained in serum. Hirakawa et al. developed an “NMR modal analysis” method [18] that enables visualization of the specific nuclear magnetic resonance (NMR) signal properties of a biological sample by measuring hydrogen atoms and analyzing them using the concepts and methods of modal analysis in vibration engineering. Data indicate that clinical specimens such as serum analyzed using this method exhibit different “NMR modes” depending on the presence of a disease or other pathological condition. As an example, there are published studies in which NMR modal analysis of serum from pancreatic cancer patients yielded results of diagnostic importance [19]. Therefore, we considered the possibility of applying our method not only to pancreatic cancer, but also to other diseases that indirectly or directly alter the physical properties of serum, specifically, the mobility of all hydrogen nuclei contained in serum. In the present study, we undertook an examination of the feasibility of NMR modal analysis for the evaluation and management of arteriosclerotic lesions associated with diabetes mellitus.

As patients with diabetes mellitus are multifactorial in terms of lifestyle, environment, and heredity [20, 21], it is difficult to group patients for studies of the specific NMR signal properties of arteriosclerotic lesions. For that reason, in the present study we utilized an animal model of diabetes with clearly identified risk factors in order to verify whether NMR modal analysis would be useful for the evaluation of arteriosclerotic lesions in humans.

Objective

We conducted an arteriosclerotic lesion model experiment. The animal group was constructed using a diabetes mouse model of hyperglycemia, a risk factor for diabetes mellitus, from an early age. This study aimed to explore the possibility of early detection and evaluation of arteriosclerotic lesions from serum samples by NMR modal analysis using a diabetes mouse model with rapid onset and progression of arteriosclerotic lesions.

Method

All mouse model protocols were in accordance with National Institute of Health guidelines. This study was approved by the Institutional Animal Care and Use Committee of our institution (permission number: 2020–094) and carried out according to the Nippon Medical School Animal Experimentation Regulations.

Breeding of experimental animals

Diabetic model mice (BKS.Cg db/db) and control mice (Jcl:ICR) were purchased from Nippon Crea Co., Ltd., Tokyo, Japan, and kept in a clean room in the Laboratory Animal Care Room of Nippon Medical School. The diet consisted of general solid food (MF from Oriental Yeast Co., Ltd., Tokyo, Japan), and the animals were allowed to eat and drink freely.

Understanding of blood glucose dynamics

Glucose tolerance testing and blood glucose monitoring were performed on a set of 16 mice (4 mice each of the Jcl:ICR, C57BL/6JJcl, BKS.Cg db/db, and BKS.Cg -/db strains) during rearing to determine their blood glucose dynamics. Glucose tolerance tests were conducted periodically from the age of 8 weeks. After fasting for 12 h from the previous day, 2 g/kg of 50% glucose was administered, and blood glucose was measured at 0 min, 30 min, 1 h, and 2 h after administration using a commercially available rapid kit for humans (Safe-ACCU, Sinocare, No.265 Guyuan Road Hi-tech Zone, Changsha, Hunan, China). The kit was unable to take measurements when blood glucose exceeded 600 mg/dL. At 8 weeks of age, the fasting blood glucose exceeded 200 mg/dL, which made measurement with the rapid kit difficult when glucose was loaded. Blood glucose was also measured sporadically, and the animals were confirmed to be hyperglycemic at each week of life. No clear sex differences in blood glucose measurements were observed at any week or in any lineage.

Weeks of sampling and sample size

Based on the hemodynamic parameters described in the previous section, we estimated 10–26 weeks to be the age at which hyperglycemia was maintained, making that an appropriate time to evaluate changes in serum data and pathological findings. Samples were taken every 4 weeks for 26 weeks, starting at 10 weeks of age for BKS.Cg db/db mice, apart from the aforementioned set of mice.

We initially chose a small sample size for the first analysis of mouse serum using NMR modal analysis. Specifically, two males and two females were selected for each 1-week-old group of BKS.Cg db/db mice. However, we later increased the sample size for 1-week-old mice with rapidly progressing lesions to better identify atherosclerotic lesions, which was the objective of our study. The total number of mice sampled was 28 BKS.Cg db/db (4 mice each at 10, 14, 22, and 26 weeks of age and the remaining 12 mice at 18 weeks of age) and 7 Jcl:ICR mice.

Sample collection and tests performed

The animals were euthanized by opening the chest and drawing cardiac blood (approximately 1 mL) after maintenance anesthesia with isoflurane to obtain adequate analgesia.

Collected blood was centrifuged, and serum was aliquoted for the biochemistry testing and NMR analysis. Samples for NMR measurement were stored at −80°C until analysis [22]. Samples for blood biochemical tests were selected from one male and one female in each group of mice. In addition to blood glucose, total cholesterol (T-Cho) and triglyceride (TG) levels were measured as serum components related to arteriosclerotic lesions. Measurements were outsourced to Oriental Yeast Co. (Shiga, Japan).

Pathological testing

After blood collection, carotid arteries were collected for histological examination by hematoxylin and eosin (H&E) staining. Carotid specimens for pathological examination could not be obtained from specimen no. 22–2 (22 weeks old) due to difficulty in identifying the carotid artery. After collection, the specimens were immediately fixed in formalin solution. Processing after embedding and H&E staining were outsourced to Genostaff Co. (Tokyo, Japan).

Free induction decay data collection for NMR measurement

Each serum sample (100 μL) was mixed with 500 μL of deuterium oxide (²H₂O) (ISOTEC, Sigma-Aldrich, St. Louis, MO, USA) and pipetted into a 5-mm (outer diameter) NMR tube (Wilmad-LabGlass, Vineland, NJ, USA) for NMR analysis [23]. Solution-state NMR analyses were performed at a proton resonance frequency of 400 MHz (9.4 Tesla) using an ECZ^™ NMR spectrometer (JEOL Ltd., Tokyo, Japan) interfaced with a probe (digital auto-tunable type [NM-03812RO5S]) and equipped with Delta^™ NMR processing and control software, version 5.3.2 (JEOL Ltd.). The field was locked to the ²H resonance of the ²H₂O solvent. One-dimensional proton NMR signals were automatically acquired at a probe temperature of 30°C using the program supplied by JEOL that supported the macro function in Delta^™. Free induction decay (FID) data were acquired using a single pulse with a 2.0-s relaxation delay between repeated pulse sequences. The strong signal arising from free water was suppressed using DANTE presaturation. Other conditions were as follows: radio-frequency pulse width, 2.93 μs; acquisition time, 1.636 s; repetition time, 3.636 s; spectral width, 10,016 Hz; number of data points, 16,384; and number of steady-state transients, 400. The FID data were saved in JEOL Delta format (JDF).

NMR data analysis

For time-frequency analysis, which analyzed how the frequency content of FIDs changed over time, short-time Fourier transform (STFT) was used. The STFT was calculated using a self-made application program (“STFT tool”) developed for MATLAB^™ (The MathWorks, Inc., Natick, MA, USA), a programming and numeric computing platform.

The FID data for each serum sample were imported into the STFT tool, and STFT processing and spectrogram calculation were performed. The spectrogram data were output as images (TIFF format) and ASCII data. The STFT and spectrogram calculations were performed using a previously reported algorithm [24].

The STFT matrix was given by X (f) = [X₁ (f) X₂ (f) X₃ (f) X_k (f)], such that the m^th element of the matrix was:

X_{m} (f) = \sum_{n = \infty}^{\infty} x (n) g (n - m R) e^{- j 2 π f n}

In this formula, g (n) represents the window function of length M, X_m (f) represents the DFT of windowed data centered about time mR, and R represents Hop size (the difference between the window length M and the overlap length L) between successive DFTs.

The number of columns in the STFT matrix was given by:

k = ⌊\frac{N_{x} - L}{M - L}⌋

In this formula, Nx represents the length of the original signal x (n), and the ⌊ ⌋ symbols denote the floor function. The number of rows in the matrix equaled NDFT, the number of DFT points, for centered and two-sided transforms and ⌊NDFT/2⌋ + 1 for one-sided transforms.

When a nonzero overlap length L was specified, it was represented by R = M-L. R varied from 0 to M − 1. f represents the index spanning and varies from 0 to (FFT block size − 1). The parameters used in this study were as follows: g hanning, L 64, R 32, FFT block size 1024 (= 1/n), and M 256. A spectrogram was generated as a visual representation of the time-frequency analysis. The data for the STFT-based spectrogram, P_SP, were calculated by the square of X(f). Each spectrogram datum in the two-dimensional matrix was read from left to right and row by row, populated in a single row. In this study, each spectrogram datum was reshaped from a 256×1024 two-dimensional matrix into a 1×262,144 single row. After combining the total rows into a separate matrix, (number of samples)×262,144, the resulting dataset was used for multivariate data analysis.

Multivariate analysis

Similar to the study of the diagnosis of pancreatic cancer from serum by applying NMR modal methods [19], pattern recognition [25] of image data was applied to clarify groups from ¹H NMR-FID spectrograms of mouse serum in this study, and principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA) were performed. To perform PCA and PLS-DA, all individual spectrogram data were imported into the Unscrambler X software (version 11; Camo Software AS, Oslo, Norway). After importation, the data were pre-processed by Standard Normal Variates transformation (Standard Normalization).

PCA is a method of dimensionality reduction of data using only explanatory variables and can be used for exploratory analyses such as identifying outliers and potential subgroups [26]. In this study, the PCA method was used to examine the relationship between the principal component score of each individual and the age in weeks and pathological findings [27]. Based on the results of the PCA, we regrouped the spectrogram data according to age in weeks and the progression of atherosclerotic lesions. We then analyzed the spectrogram data using PLS-DA [28–30] to visualize the presence or absence of differences between the regrouped mice and the time-frequency regions of the NMR signals associated with the differences.

The NMR signal regions associated with the differences between groups were considered the key variables, and to visualize these regions, the correlation loading values for each factor were plotted on the spectrogram. This was referred to as a “correlation loading plot”.

The score plot was displayed using the Unscrambler drawing function, and the correlation loading plot was displayed using a program written for MATLAB. Variables with an absolute value ≥0.4 as the loading value were considered relevant.

A graphical flow chart from the animal experiments to the multivariate analysis performed in this study is shown as supporting information (S1 and S2 Figs).

Results

Pathological results

Carotid artery pathology was observed, and Table 1 lists the lesions observed at each week for each mouse of the Jcl:ICR and BKS.Cg db/db strains, except for mouse ID.B22-2, from which carotid artery samples could not be obtained at necropsy. For Jcl:ICR, observations were only made at 18 weeks of age, and although all mice exhibited mild diffuse intimal thickening, no atheromatous plaques were observed. In the BKS.Cg db/db mice, diffuse intimal thickening, an indicator of moderate atherosclerotic lesions, was observed in 3 of 4 mice at 10 weeks of age and in all mice after 14 weeks of age. No atherosclerotic plaques were observed in any of the mice at 10 or 14 weeks of age. At 18 weeks of age, plaques were present in 5 of 6 males and 2 of 6 females. At 22 and 26 weeks, the presence of atherosclerotic plaques was confirmed in all observed mice.

Table 1. Summary of Jcl:ICR and BKS.Cg db/db carotid artery lesion evaluation.

	Week	Sex	Subject ID	No lesion	Diffuse intimal thickening	Atheroma	Severity
Jcl:ICR	18	male (n = 3)	J18-1,2,3	0	3	0	Moderate
Jcl:ICR	18	female (n = 4)	J18-4,5,6,7	0	4	0	Moderate
BKS	10	male (n = 2)	B10-1	0	1	0	Moderate
		male (n = 2)	B10-2	1	0	0	Mild
		female (n = 2)	B10-3, 4	0	2	0	Moderate
	14	male (n = 2)	B14-1,2	0	2	0	Moderate
	14	female (n = 2)	B14-3,4	0	2	0	Moderate
	18	male (n = 6)	B18-3	0	1	0	Moderate
		male (n = 6)	B18-1,2,4,5,6	0	5	5	Severe
		female (n = 6)	B18-8,9,10,11	0	4	0	Moderate
		female (n = 6)	B18-7,12	0	2	2	Severe
	22	male (n = 2)	B22-1	0	1	1	Severe
			B22-2	none	none	none
			B22-2	none	none	none
		female (n = 2)	B22-3	0	1	0	Moderate
		female (n = 2)	B22-4	0	1	1	Severe
	26	male (n = 2)	B26-1,2	0	2	2	Severe
	26	female (n = 2)	B26-3,4	0	2	2	Severe

Open in a new tab

The presence or absence of 3 items of carotid artery pathological findings in the two mouse strains, Jcl:ICR and BKS.Cg db/db, was tabulated for each week of age. The degree of progression of atherosclerotic lesions was judged as moderate for diffuse intimal thickening, severely advanced for atheromatous plaques, and mild for none of the above.

In the Jcl:ICR sample, although diffuse intimal thickening was indicated to some extent, none of the samples exhibited marked thickening, and none had atheromatous plaques.

In BKS.Cg db/db samples, mice already exhibited intimal thickening at 10 weeks of age, and many more mice developed atheromaatous plaques after 18 weeks of age. In BKS.Cg db/db, individuals with atheromatous plaques appeared after 18 weeks of age. The proportion increased stepwise as age increased. In mouse B22-2 only, the carotid artery could not be identified and collected for histological examination.

Representative H&E staining of the carotid arteries of BKS.Cg db/db mice is shown in Fig 1(A) B10-1, Fig 1(B) B14-4, and Fig 1(C) B18-5. Fig 1D shows a photograph of H&E staining of J18-1. BKS.Cg db/db mice exhibited clear exacerbation of the degree of intimal thickening at the ages of 10, 14, and 18 weeks, as atherosclerosis progressed over a short period. At 18 weeks of age, BKS.Cg db/db mice exhibited a greater degree of intimal thickening than Jcl:ICR mice at 18 weeks of age, and the presence of atheromatous plaques was evident in more than half of the animals, indicating that atherosclerosis progressed rapidly at this age. At 18 weeks of age, there was a difference between males and females in terms of the presence or absence of atheromatous plaques.

Blood biochemistry tests

Table 2 shows the results of biochemical tests performed on two mice during each week. Blood glucose levels exceeded 600 mg/dL in all BKS.Cg db/db mice. Conversely, T-Cho values were generally slightly higher in BKS.Cg db/db mice than in Jcl:ICR mice, but they did not differ by week. Neutral fat values varied depending on the previous feeding condition, and under free-feeding conditions, there was a large variation among individuals, and no strain or week-of-age–related characteristics were confirmed.

Table 2. Results of biochemical tests performed on two BKS.Cg db/db mice at each week of age.

	No.	BS (mg/dL)	T-Cho (mg/dL)	TG (mg/dL)
BKS.Cg db/db 10W	B10-1	846	146	328
BKS.Cg db/db 10W	B10-3	818	118	146
BKS.Cg db/db 14W	B14-1	681	131	196
BKS.Cg db/db 14W	B14-3	796	155	217
BKS.Cg db/db 18W	B18-2	647	132	86
BKS.Cg db/db 18W	B18-9	766	134	210
BKS.Cg db/db 22W	B22-2	838	154	182
BKS.Cg db/db 22W	B22-3	1004	92	90
BKS.Cg db/db 26W	B26-1	832	141	122
BKS.Cg db/db 26W	B26-4	867	107	94
Jcl:ICR 18W	J18-2	127	130	207
Jcl:ICR 18W	J18-6	84	86	56

Open in a new tab

An elevated blood glucose level exceeding 600 mg/dL was observed in both weeks, whereas total cholesterol was only slightly higher than in Jcl:ICR mice.

Triglyceride levels varied depending on the previous feeding situation in both the Jcl:ICR and BKS.Cg db/db groups.

NMR modal analysis

Fig 2 shows a representative spectrogram of specimen B18-1. Although all spectrogram images were aligned and visually observed and compared with respect to strain, age, and pathological findings, identifying individual characteristics was difficult.

The PCA on spectrogram data of all mouse serum samples showed that the BKS and control groups clustered in the PC1/PC2 score plots (data not shown).

The results of PCA on the spectrogram data for serum samples from all BKS.Cg db/db mice are shown in Fig 3. In this figure, the PC1/PC5 score plot (Fig 3A) showed that groups B10 and B14 formed one loose cluster with similar PC1 scores. PC1 scores were organized as follows: B26 group > B22 group > B10 and B14 groups. In contrast, the PC1 scores of the B18 group varied widely from individual to individual, making it difficult to characterize the B18 group as a single population. On the score plot, we observed two subgroups: six mice with negative PC1 scores, similar to the B10 and B14 groups (B18-2, B18-3, B18-4, B18-6, B18-9, and B18-11), and six mice with positive PC1 scores, similar to the B22 and B26 groups (B18-1, B18-5, B18-7, B18-8, B18-10, and B18-12). Fig 3B shows a correlation loading plot of PC1. The regions of different positive and negative loading values are clearly displayed in the NMR signal-derived regions observed in the spectrogram of Fig 2, indicating that the PC1 scores reflect differences in the time-frequency characteristics of the NMR signals of the samples.

The results of PLS-DA in groups B10 and B14 and groups B22 and B26, which had a tendency to form clusters, are shown in Fig 4. In the Factor-1/Factor-2 score plot (Fig 4A), groups B10 and B14 and groups B22 and B26 were roughly divided by Factor-1 into clear clusters, with the B10 and B14 groups forming a cluster with similar scores and the B22 and B26 groups, except for B26-1 and B26-2, also forming a loose cluster. Fig 4B displays a correlation loading plot of Factor-1, showing the presence of variables important for the separation of the two groups on the time-frequency domain of the NMR signal. Because PLS-DA was used as a 2-class classifier in the present study, Factor-1 in Fig 4 is interpreted as an axis that reflects group information; axes from Factor-2 onward are interpreted as axes that maximize the separation on Factor-1 and make it visible on the plot. This is also the case for Figs 5 through 7. This interpretation applies to Figs 5–7 as well.

Fig 7 — (A) Factor-1/Factor-3 score plot, with Factor-1 dichotomized between 18- and 10- and 14-week-old mice. (B) Factor-1 loading diagram. Some of the loading values seem to be similar to those in Fig 5B.

Fig 5 shows the results of the PLS-DA performed on the aforementioned PCA, distinguishing six mice from group B18 with negative PC1 scores, similar to groups B10 and B14, and mice from group B18 with positive PC1 scores, similar to groups B22 and B26. Negative value samples are denoted as 18W A, and positive value samples are denoted as 18W B. In the Factor-1/Factor-2 score plot (Fig 5A), the results are dichotomized by Factor-1. Fig 5B is a correlation loading plot for Factor-1, showing the presence of variables important for the separation of the two groups on the time-frequency domain of the NMR signal.

PLS-DA was performed by combining groups B10, B14, and B18 A and groups B22, B26, and B18 B, respectively, under the assumption that the total sample was dichotomized within the 18-week-old group. The results are shown in Fig 6, in which groups B10, B14, and B18 A are labeled as precursor groups and groups B22, B26, and B18 B as progression groups. In the Factor-1/Factor-2 score plot (Fig 6A), the results were dichotomized by Factor-1. Fig 6B is a correlation loading plot of Factor-1, showing the presence of variables important for the separation of the two groups on the time-frequency domain of the NMR signal.

As shown in Fig 6, although six mice from groups B10, B14, and B18 A grouped together as a precursor group, as in the first PCA method, the score values of these groups differed slightly. Fig 7 displays the results of PLS-DA with respect to these data. In the Factor-1/Factor-3 score plot (Fig 7A), Factor-1 dichotomizes the two groups into B10, B14, and B18 A groups. Fig 7B is a correlation loading plot of Factor-1, showing the presence of variables important for the separation of the two groups on the time-frequency domain of the NMR signal.

Discussion

The results of this study showed that NMR modal properties of serum samples resulting from analysis of NMR-FID signals were associated with arteriosclerotic lesions. In addition, this study successfully showed that NMR modal properties can vary as arteriosclerotic lesions progress. It can be pointed out that one of the factors contributing to this success was the successful completion of animal experiments conducted under well-controlled conditions, and with the exception of one individual, pathology experiments on mouse carotid arteries were also successful. In fact, it has been pointed out that the first animal experiment is a very important element in the interpretation of the results of multivariate analysis in the step of chemometric studies [31].

Based on the score plots of all PCA data for BKS.Cg db/db mice, all groups (except B18) were divided into a precursor group and progression group by PC-1. In contrast, the data plots for group B18 were scattered and bifurcated into precursor and progression groups, as the arteriosclerotic lesions progressed particularly rapidly. If the scatter of the score plots in Fig 3 reflected differences in week age, a distribution by age group would be expected, but the actual results are contrary to this. Since PCA is an analysis that uses only explanatory variables (in this experiment, the serum spectrogram data) for dimensionality reduction [26], it was assumed that this result could be interpreted as a visualization of the change in serum status before and after 18 weeks of age.

Based on the results of PLS-DA (Fig 6A), the data plots for group B18 were ultimately divided into a precursor group and a progression group. Careful pathological examination showed that some samples from group B18 exhibited evidence of atheromatous plaques; part of these samples was classified into the precursor group. As shown in Fig 7A, the PLS-DA results for groups B10, B14, and B18 A showed that the samples were divided into these groups by Factor-1. The PLS-DA score plot for this precursor group shown in Fig 7A cannot be explained by age-related changes alone. In this study, time-trend changes in carotid artery pathology were observed, which, as noted above, showed rapid atherosclerotic lesion progression at 18 weeks. Considering the characteristics of the distribution of the B18 group shown on some score plots, we interpret this as suggesting that the clustering and separation of samples shown in the multivariate score plots in this study may be related to the progression of atherosclerotic lesions, though there is a certain degree of influence of differences in age.

In addition, the time-frequency domains of the variables valid for discrimination shown in the loading plots in Figs 3–7 corresponded roughly to the signal domains in the serum spectrograms in Fig 2. These results suggest that the progression of arteriosclerotic lesions was reflected in the serum NMR modal properties.

NMR, which enables analyses of the structure and energy state of matter at the atomic level, is an analytical method that examines interactions between nuclei and radiofrequency waves [32]. Medical applications of NMR include relaxometry and spectroscopy. Relaxometry measures and compares the relaxation time (time-series data analysis) of an FID signal, such as the basic technique of magnetic resonance imaging, which is widely used in clinical medicine. Spectroscopy is applied in the comprehensive chemical analysis of metabolites [33–37] by measuring and comparing spectra (frequency analysis). Spectroscopy is a basic research tool in the field of metabolomics, and several reports [38–42] have suggested its suitability for use in clinical diagnosis. The NMR modal analysis technique used in this study is based on the concept of “modal analysis” in vibration engineering [43, 44], which analyzes FID signal properties.

The present data suggest that this NMR technique is highly suitable for analyzing biomedical samples. Serum, which was targeted in this study, is a complex mixture containing considerable amounts of albumin in addition to numerous metabolites and small ionic molecules. Hydrogen atoms contained in water and other molecules in serum play significant roles in interactions within and between molecules. As such, spectrograms calculated from FID were prepared to visualize the properties of the NMR mode for an entire sample, which was determined by the interactions with protons within and between molecules. The correlation loading plots indicate which parts of the NMR modal properties are valid for discrimination. These plots can be interpreted as presenting a visualization of the parts of the overall properties of the serum sample that are relevant to the characteristics of the mouse model of atherosclerosis. Therefore, the results of this study suggest that specific NMR modal properties of serum are associated with the onset of arteriosclerosis due to hyperglycemia and that such properties can be identified using this approach. NMR metabolomics experiments focus on low molecular weight substances and obtain their spectra [45]. On the other hand, in NMR mode analysis, all molecular species including protons of up to high molecular weight substances such as serum proteins in addition to low molecular weight metabolites are measured by NMR in a non-selective and comprehensive manner to obtain spectrograms that reflect the state of all hydrogen atoms in the serum sample. This allows us to interpret that biochemical changes due to disease or pathological abnormalities may be reflected in not only static but also dynamic changes of protons in all components in serum (from low to high molecular weight substances and free water). In this study, it should be noted that, since atherosclerotic lesions are associated with certain biochemical changes in serum (inflammatory markers and hormones), the modal properties of this serum NMR should be considered to include inflammatory markers and hormonal changes associated with the progression of atherosclerotic lesions.

This study had some limitations. First, experimental animals were used in this study, so no human blood samples were analyzed. In addition, arteriosclerosis can be caused by multiple factors other than diabetes, such as hypertension and lipid abnormalities, but the mice used in this study had only persistent hyperglycemia. It is also known that the risk of cardiac and renal events increases as the number of risk factors increases and arteriosclerotic lesions progress [8–10, 16, 17, 46, 47]. Second, there was some discordance between the multivariate analysis results and pathological findings of BKS.Cg db/db mice based on differences in the biological reactions in individual mice. The biological reactions may partly reflect the serum of mice with atherosclerosis. On the other hand, in 18-week-old mice with rapidly progressive atherosclerosis, the reaction may be associated with pathological progression. Unfortunately, our exploratory studies are limited in their ability to elucidate the true nature of this biological response, and it will be the subject of future research. These limitations indicate that our results cannot be immediately extrapolated to humans.

Despite the above-mentioned limitations of this study, the arteriosclerosis model experiments using BKS.Cg db/db mice, which have persistent hyperglycemia, were successful. Continued hyperglycemia is diagnosed as diabetes in human clinical medicine, but there are no clear diagnostic criteria for diabetes in laboratory animals [48]. Thus, both glucose tolerance tests and occasional blood glucose measurements [49, 50] were performed in the present study, confirming that the mice had hyperglycemia before sexual maturity, which is considered indicative of diabetes in humans. Based on these findings, we consider that it may be worth exploring the possibility that arteriosclerosis, a risk factor for cardiovascular events in diabetic patients, can be assessed using NMR modal analysis of serum samples.

Conclusions

The results of this study suggest that NMR modal analysis of serum samples obtained from mice with persistent hyperglycemia can be used to monitor the development of atherosclerosis. The mice used in the present study exhibited severe atherosclerotic lesions at 18 weeks of age, and serum samples from the borderline weeks showed differences in NMR modal data. We expect that further studies and appropriate modeling will enable clinical examination of the progression of diabetes-related atherosclerotic lesions using blood samples.

Supporting information

S1 Fig. Graphical flow chart from the animal experiments to the multivariate analysis.

In this figure, the results of NMR analysis of blood samples collected after animal experiments in groups A (n = 10) and B (n = 10) are shown. In total, 20 ¹H-NMR FID signal data were acquired, which were short-time Fourier transformed [51], and 20 spectrograms were calculated. For multivariate analysis, a process was performed to obtain a data matrix from the 10 spectrograms. From this data matrix, an exploratory analysis was performed using principal component analysis [26] for the identification of outliers and potential subgroups of animals in groups A and B. From this figure, it was possible to identify subgroups that were not initially anticipated, so PLS-DA [52, 53] was performed only with data that corresponded to subgroups in the same data matrix to analyze whether subgroups could be identified. In both principal component analysis and PLS-DA, the variables that were important for the identification of groups in each analysis were plotted as corresponding score plots. This score plot has the same time and frequency axes as the spectrogram and thus can be compared with the spectrogram. The illustration of the mouse and the NMR equipment in this figure was obtained from Research Net, a website where research illustration material is available free of charge (https://www.wdb.com/kenq/illust ^© WDB Co., Ltd., Tokyo, Japan).

References (cited only in the Supporting Information)

51. Wacker M, Witte H. Time-frequency techniques in biomedical signal analysis. a tutorial review of similarities and differences. Methods Inf Med. 2013;52:279–296.

52. Barker M, Rayens W. Partial least squares for discrimination. J Chemo. 2003;17:166–173.

53. Sylvie Chevallier S, Bertrand D, Kohler A, Courcoux P. Application of PLS-DA in multivariate image analysis. J Chemo. 2006;20: 221–229.

(TIF)

pone.0299641.s001.tif^{(464KB, tif)}

S2 Fig. Graphical flow chart of matrixing of spectrogram data.

The process surrounded by the green box in S1 Fig is described in detail. For multivariate analysis of spectrogram data, a 256 × 1024 spectrogram (a) with frequency on the horizontal axis and time on the vertical axis was divided into 1024 single rows (b). All rows were re-arranged into a single row, and 1 × 262,144 single rows were reshaped (c). Since a single row represents spectrogram data for one individual (d), the single rows of spectrogram data for all individuals needed for multivariate analysis are combined to create a data matrix (e).

(TIF)

pone.0299641.s002.tif^{(563.4KB, tif)}

S1 Data

(ZIP)

pone.0299641.s003.zip^{(7.7MB, zip)}

Acknowledgments

We thank Haruo Hashimoto for guidance in the animal experiments, Kaori Okihara and Tomoko Konta for assistance with NMR measurements, and Hironobu Yoshimura, Naohito Ogiso, and Tomoki Nakao of JEOL, Ltd. (Tokyo, Japan) for providing technical assistance with the NMR data analysis.

Data Availability

All relevant data are within the paper and its Supporting information files.

Funding Statement

The author(s) received no specific funding for this work.

References

1.Grundy SM, Benjamin IJ, Burke GL, Chait A, Eckel RH, Howard BV, et al. Diabetes and cardiovascular disease: a statement for healthcare professionals from the American Heart Association. Circulation. 1999;100: 1134–1146. doi: 10.1161/01.cir.100.10.1134 [DOI] [PubMed] [Google Scholar]
2.Gæde P, Lund-Andersen H, Parving H-H, Pedersen O. Effect of a multifactorial intervention on mortality in type 2 diabetes. N Engl J Med. 2008;358: 580–591. doi: 10.1056/NEJMoa0706245 [DOI] [PubMed] [Google Scholar]
3.Zureik M, Bureau JM, Temmar M, Adamopoulos C, Courbon D, Bean K, et al. Echogenic carotid plaques are associated with aortic arterial stiffness in subjects with subclinical carotid atherosclerosis. Hypertension. 2004;41: 519–527. [DOI] [PubMed] [Google Scholar]
4.Antiga L, Wasserman BA, Steinman DA. On the overestimation of early wall thickening at the carotid bulb by black blood MRI, with implications for coronary and vulnerable plaque imaging. Magn Reson Med. 2008;60: 1020–1028. doi: 10.1002/mrm.21758 [DOI] [PubMed] [Google Scholar]
5.Bartlett ES, Walters TD, Symons SP, Fox AJ. Quantification of carotid stenosis on CT angiography. Am J Neuroradiol. 2006;27: 13–19. [PMC free article] [PubMed] [Google Scholar]
6.Ueda M, Takabe M, Sawada T, Urai S, Hashimoto N, Ohara T. Comparative study of vascular endothelial function tests (FMD, RHI) in patients with type 2 diabetes. J Diabetes. 2017;60: 370–377. [Google Scholar]
7.Fukui S. Tests to be performed for prevention of arteriosclerosis in physical health checkups. Ningen Dokku. 2016;30: 809–821. [Google Scholar]
8.The Japan Diabetes Society, Guidelines for Diabetes Care 2019. Nankkodo. 2019: 201–218.
9.The Japanese Society of Hypertension, Committee for the Preparation of Guidelines for the Treatment of Hypertension, Guidelines for the Treatment of Hypertension 2019, Japanese Society of Hypertension, 2019: 25–30.
10.Okada E, Oida K, Tada H, Asazuma K, Eguchi K, Tohda G, et al. Hyperhomocysteinemia is a risk factor for coronary arteriosclerosis in Japanese patients with type 2 diabetes. Diabetes Care. 1999;22: 484–490. doi: 10.2337/diacare.22.3.484 [DOI] [PubMed] [Google Scholar]
11.Lopes-Virella MF, Virella G. Cytokines, modified lipoproteins, and arteriosclerosis in diabetes. Diabetes. 1996;45: 40–45. doi: 10.2337/diab.45.3.s40 [DOI] [PubMed] [Google Scholar]
12.Lundberg AM, Hansson GK. Innate immune signals in atherosclerosis. Clin Immunol. 2010;134: 5–24. doi: 10.1016/j.clim.2009.07.016 [DOI] [PubMed] [Google Scholar]
13.López-Tinoco C, Roca M, Fernández-Deudero A, García-Valero A, Bugatto F, Aguilar-Diosdado M, et al. Cytokine profile, metabolic syndrome and cardiovascular disease risk in women with late-onset gestational diabetes mellitus. Cytokine. 2021;58: 14–19. [DOI] [PubMed] [Google Scholar]
14.Severino P, D’Amato A, Netti L, Pucci M, Infusino F, Maestrini V, et al. Myocardial ischemia and diabetes mellitus: role of oxidative stress in the connection between cardiac metabolism and coronary blood flow. J Diabetes Res. 2019;2019: 9489826. doi: 10.1155/2019/9489826 [DOI] [PMC free article] [PubMed] [Google Scholar]
15.Liguori A, Abete P, Hayden JM, Cacciatore F, Rengo F, Ambrosio G, et al. Effect of glycaemic control and age on low-density lipoprotein susceptibility to oxidation in diabetes mellitus type 1. Eur Heart J. 2001;22: 2075–2084. doi: 10.1053/euhj.2001.2655 [DOI] [PubMed] [Google Scholar]
16.Bierhaus A, Hofmann MA, Ziegler R, Nawroth PP. AGEs and their interaction with AGE-receptors in vascular disease and diabetes mellitus. I. The AGE concept. Cardiovasc Res. 1998;37: 586–600. doi: 10.1016/s0008-6363(97)00233-2 [DOI] [PubMed] [Google Scholar]
17.Singh RB, Mengi SA, Xu Y-J, Arneja AS, Dhalla NS. Pathogenesis of atherosclerosis: A multifactorial process. Exp Clin Cardiol. 2002;7: 40. [PMC free article] [PubMed] [Google Scholar]
18.Hirakawa K, Koike K, Kanawaku Y, Moriyama T, Sato N, Suzuki T, et al. Short-time Fourier transform of free induction decays for the analysis of serum using proton nuclear magnetic resonance. J Oleo Sci. 2019;68: 369–378. doi: 10.5650/jos.ess18212 [DOI] [PubMed] [Google Scholar]
19.Sato A, Masui T, Yogo A, Ito T, Hirakawa K, Kanawaku Y, et al. Time-frequency analysis of serum with proton nuclear magnetic resonance for diagnosis of pancreatic cancer. Sci Rep. 2020;10: 21941. doi: 10.1038/s41598-020-79087-3 [DOI] [PMC free article] [PubMed] [Google Scholar]
20.Zimmet P, Alberti KGMM, Shaw J. Global and societal implications of the diabetes epidemic. Nature. 2001;414: 782–787. doi: 10.1038/414782a [DOI] [PubMed] [Google Scholar]
21.Ray NF, Thamer M, Gardner E, Chan JK. Economic consequences of diabetes mellitus in the U.S. in 1997. American Diabetes Association. Diabetes Care. 1998;21: 296–309. [DOI] [PubMed] [Google Scholar]
22.Zheng H, Zheng Y, Zhao L, Chen M, Bai G, Hu Y, et al. Cognitive decline in type 2 diabetic db/db mice may be associated with brain region-specific metabolic disorders. Biochim Biophys Acta Mol Basis Dis. 2017;1863: 266–273. doi: 10.1016/j.bbadis.2016.11.003 [DOI] [PubMed] [Google Scholar]
23.Tajima T, Hirakawa K, Kawaguchi H, Sakamoto A. Proton nuclear magnetic resonance and pattern recognition analysis of liver extracts from rats under different anesthetics. BMC Med Imaging. 2012;12: 28. doi: 10.1186/1471-2342-12-28 [DOI] [PMC free article] [PubMed] [Google Scholar]
24.Short-Time Fourier Transform, The MathWorks, Inc. [cited 24 Aug 2023]. https://jp.mathworks.com/help/signal/ref/stft.html?lang=en
25.Amigo JM, Babamoradi H, Elcoroaristizabal S. Hyperspectral image analysis. A tutorial. Anal Chim Acta. 2015:896:34–51. doi: 10.1016/j.aca.2015.09.030 [DOI] [PubMed] [Google Scholar]
26.Wold S, Esbensen K, Geladi P. Principal component analysis. Chemometr Intell Lab Syst. 1987; 2:37–52. [Google Scholar]
27.Kawaguchi H, Hirakawa K, Miyauchi K, Koike K, Ohno Y, Sakamoto A. Pattern recognition analysis of proton nuclear magnetic resonance spectra of brain tissue extracts from rats anesthetized with propofol or isoflurane. PLoS One. 2010;5: e11172. doi: 10.1371/journal.pone.0011172 [DOI] [PMC free article] [PubMed] [Google Scholar]
28.Goodpaster AM, Kennedy MA. Quantification and statistical significance analysis of group separation in NMR-based metabolomics studies. Chemometr Intell Lab Syst. 2011;109: 162–170. [DOI] [PMC free article] [PubMed] [Google Scholar]
29.Dickens AM, Larkin JR, Davis BG, Griffin JL, Claridge TD, Sibson NR, et al. NMR-based metabolomics separates the distinct stages of disease in a chronic relapsing model of multiple sclerosis. J Neuroimmune Pharmacol. 2015;10: 435–444. doi: 10.1007/s11481-015-9622-0 [DOI] [PubMed] [Google Scholar]
30.Asano T, Hirakawa K, Koike K, Ohno Y, Fujino O. Visualization of different characteristics of cerebrospinal fluid with acute encephalopathy and febrile seizures using pattern recognition analysis of 1H NMR. Pediatr Res. 2015;77: 70–74. doi: 10.1038/pr.2014.141 [DOI] [PubMed] [Google Scholar]
31.Ebbels TMD, Cavill R. Bioinformatic methods in NMR-based metabolic profiling. Prog Nucl Magn Reson Spectrosc. 2009;55:361–374. [Google Scholar]
32.de Graaf RA. In Vivo NMR Spectroscopy: Principles and Techniques. Chichester, England. John Wiley & Sons; 2018. pp. 15–17. [Google Scholar]
33.Jung JY, Lee HS, Kang DG, Kim NS, Cha MH, Bang OS, et al. 1H-NMR-based metabolomics study of cerebral infarction. Stroke. 2011;42: 1282–1288. doi: 10.1161/STROKEAHA.110.598789 [DOI] [PubMed] [Google Scholar]
34.Okuda T, Hirakawa K, Hakan O, Ohno Y, Hyakusoku H. Nuclear magnetic resonance spectroscopy as a useful tool for ingredient analysis of synthetic materials. Plast Reconstr Surg. 2012;129: 197e–198e. doi: 10.1097/PRS.0b013e3182365d00 [DOI] [PubMed] [Google Scholar]
35.Sinclair AJ, Viant MR, Ball AK, Burdon MA, Walker EA, Stewart PM, et al. NMR-based metabolomic analysis of cerebrospinal fluid and serum in neurological diseases—a diagnostic tool? NMR Biomed. 2010;23: 123–132. doi: 10.1002/nbm.1428 [DOI] [PubMed] [Google Scholar]
36.Mora-Ortiz M, Nunez Ramos P, Oregioni A, Claus SP. NMR metabolomics identifies over 60 biomarkers associated with type II diabetes impairment in db/db mice. Metabolomics. 2019;15: 89. doi: 10.1007/s11306-019-1548-8 [DOI] [PMC free article] [PubMed] [Google Scholar]
37.Khattri RB, Thome T, Ryan TE. Tissue-specific (1)H-NMR metabolomic profiling in mice with adenine-induced chronic kidney disease. Metabolites. 2021;11: 45. doi: 10.3390/metabo11010045 [DOI] [PMC free article] [PubMed] [Google Scholar]
38.Graham SF, Holscher C, McClean P, Elliott CT, Green BD. 1H NMR metabolomics investigation of an Alzheimer’s disease (AD) mouse model pinpoints important biochemical disturbances in brain and plasma. Metabolomics. 2013;9: 974–983. [Google Scholar]
39.Emwas A-HM, Salek RM, Griffin JL, Merzaban J. NMR-based metabolomics in human disease diagnosis: applications, limitations, and recommendations. Metabolomics. 2013;9: 1048–1072. [Google Scholar]
40.Kromke M, Palomino-Schätzlein M, Mayer H, Pfeffer S, Pineda-Lucena A, Luy B, et al. Profiling human blood serum metabolites by nuclear magnetic resonance spectroscopy: a comprehensive tool for the evaluation of hemodialysis efficiency. Transl Res. 2016;171: 71–82.e71-79. doi: 10.1016/j.trsl.2016.02.003 [DOI] [PubMed] [Google Scholar]
41.Verwaest KA, Vu TN, Laukens K, Clemens LE, Nguyen HP, Van Gasse B, et al. (1)H NMR based metabolomics of CSF and blood serum: a metabolic profile for a transgenic rat model of Huntington disease. Biochim Biophys Acta. 2011;1812: 1371–1379. doi: 10.1016/j.bbadis.2011.08.001 [DOI] [PubMed] [Google Scholar]
42.Hatazawa J. Clinical application of proton magnetic resonance spectroscopy in neurological diseases. AJNR Am J Neuroradiol. 1996;17: 737–747.8730195 [Google Scholar]
43.Yang Y, Nagarajaiah S. Time-frequency blind source separation using independent component analysis for output-only modal identification of highly damped structures. J Struct Eng. 2013;139: 1780–1793. [Google Scholar]
44.Perez-Ramirez CA, Amezquita-Sanchez JP, Adeli H, Valtierra-Rodriguez M, Romero-Troncoso RdJ, Dominguez-Gonzalez A, et al. Time-frequency techniques for modal parameters identification of civil structures from acquired dynamic signals. J Vibroengineering. 2016;18: 3164–3185. [Google Scholar]
45.Dumez JN. NMR methods for the analysis of mixtures. Chem Commun (Camb). 2022; 58:13855–13872. doi: 10.1039/d2cc05053f [DOI] [PMC free article] [PubMed] [Google Scholar]
46.Beckman JA, Creager MA, Libby P. Diabetes and atherosclerosis. J Am Med Assoc. 2002;297: 2570–2581. [DOI] [PubMed] [Google Scholar]
47.Jiang G, Luk AOY, Tam CHT, Xie F, Carstensen B, Lau ESH, et al. Progression of diabetic kidney disease and trajectory of kidney function decline in Chinese patients with type 2 diabetes. Kidney Int. 2019;95: 178–187. doi: 10.1016/j.kint.2018.08.026 [DOI] [PubMed] [Google Scholar]
48.Yagihashi S, Terauchi Y, Mori Y, Ikegami H, Satoh Y. Diagnostic criteria of diabetes in laboratory animals. Journal of the Japan Diabetes Society. 2010;53: 379–384 (Japanese). [cited 24 Aug 2023]. Available from: https://www.jstage.jst.go.jp/article/tonyobyo/53/5/53_5_379/_pdf [Google Scholar]
49.Shinohara M, Masada T, Oikawa T, Masuyama T, Takahashi T, Katsuta Y, et al. Diabetes mellitus in a female SDT rat model of spontaneous type 2 diabetes mellitus. Diabetes. 2004;47: 111–116. [Google Scholar]
50.2003 Laboratory Animal Data Collection, Japan SLC, Inc. (Shizuoka, Japan) [cited 24 Aug 2023]. http://www.jslc.co.jp/pdf/mouse/019_akita.pdf.

PLoS One. 2024 Mar 8;19(3):e0299641. doi: 10.1371/journal.pone.0299641.r001

Author response to previous submission

9 Sep 2023

Attachment

Submitted filename: ResponseToReviewers.docx

pone.0299641.s004.docx^{(119.1KB, docx)}

PLoS One. doi: 10.1371/journal.pone.0299641.r002

Decision Letter 0

Jorddy Neves Cruz

12 Dec 2023

PONE-D-23-28528Time-frequency analysis reveals an association between the specific nuclear magnetic resonance (NMR) signal properties of serum samples and arteriosclerotic lesion progression in a diabetes mouse modelPLOS ONE

Dear Dr. Kanawaku,

Thank you for submitting your manuscript to PLOS ONE. After careful consideration, we feel that it has merit but does not fully meet PLOS ONE’s publication criteria as it currently stands. Therefore, we invite you to submit a revised version of the manuscript that addresses the points raised during the review process.

==============================

ACADEMIC EDITOR: Dear authors, the reviewers raised important questions about the manuscript.

You must provide a point-by-point answer to all questions; all changes to the manuscript must be marked in yellow.

==============================

Please submit your revised manuscript by Jan 26 2024 11:59PM. If you will need more time than this to complete your revisions, please reply to this message or contact the journal office at plosone@plos.org. When you're ready to submit your revision, log on to https://www.editorialmanager.com/pone/ and select the 'Submissions Needing Revision' folder to locate your manuscript file.

Please include the following items when submitting your revised manuscript:

A rebuttal letter that responds to each point raised by the academic editor and reviewer(s). You should upload this letter as a separate file labeled 'Response to Reviewers'.
A marked-up copy of your manuscript that highlights changes made to the original version. You should upload this as a separate file labeled 'Revised Manuscript with Track Changes'.
An unmarked version of your revised paper without tracked changes. You should upload this as a separate file labeled 'Manuscript'.

If you would like to make changes to your financial disclosure, please include your updated statement in your cover letter. Guidelines for resubmitting your figure files are available below the reviewer comments at the end of this letter.

If applicable, we recommend that you deposit your laboratory protocols in protocols.io to enhance the reproducibility of your results. Protocols.io assigns your protocol its own identifier (DOI) so that it can be cited independently in the future. For instructions see: https://journals.plos.org/plosone/s/submission-guidelines#loc-laboratory-protocols. Additionally, PLOS ONE offers an option for publishing peer-reviewed Lab Protocol articles, which describe protocols hosted on protocols.io. Read more information on sharing protocols at https://plos.org/protocols?utm_medium=editorial-email&utm_source=authorletters&utm_campaign=protocols.

We look forward to receiving your revised manuscript.

Kind regards,

Jorddy Neves Cruz

Academic Editor

PLOS ONE

Journal requirements:

When submitting your revision, we need you to address these additional requirements.

1. Please ensure that your manuscript meets PLOS ONE's style requirements, including those for file naming. The PLOS ONE style templates can be found at

https://journals.plos.org/plosone/s/file?id=wjVg/PLOSOne_formatting_sample_main_body.pdf and

https://journals.plos.org/plosone/s/file?id=ba62/PLOSOne_formatting_sample_title_authors_affiliations.pdf

2. Note from Emily Chenette, Editor in Chief of PLOS ONE, and Iain Hrynaszkiewicz, Director of Open Research Solutions at PLOS: Did you know that depositing data in a repository is associated with up to a 25% citation advantage (https://doi.org/10.1371/journal.pone.0230416)? If you’ve not already done so, consider depositing your raw data in a repository to ensure your work is read, appreciated and cited by the largest possible audience. You’ll also earn an Accessible Data icon on your published paper if you deposit your data in any participating repository (https://plos.org/open-science/open-data/#accessible-data).

3. We note that you have included the phrase “data not shown” in your manuscript. Unfortunately, this does not meet our data sharing requirements. PLOS does not permit references to inaccessible data. We require that authors provide all relevant data within the paper, Supporting Information files, or in an acceptable, public repository. Please add a citation to support this phrase or upload the data that corresponds with these findings to a stable repository (such as Figshare or Dryad) and provide and URLs, DOIs, or accession numbers that may be used to access these data. Or, if the data are not a core part of the research being presented in your study, we ask that you remove the phrase that refers to these data.

4. Please include captions for your Supporting Information files at the end of your manuscript, and update any in-text citations to match accordingly. Please see our Supporting Information guidelines for more information: http://journals.plos.org/plosone/s/supporting-information.

Additional Editor Comments:

Dear authors, the reviewers raised important questions about the manuscript.

You must provide a point-by-point answer to all questions; all changes to the manuscript must be marked in yellow.

[Note: HTML markup is below. Please do not edit.]

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. Is the manuscript technically sound, and do the data support the conclusions?

The manuscript must describe a technically sound piece of scientific research with data that supports the conclusions. Experiments must have been conducted rigorously, with appropriate controls, replication, and sample sizes. The conclusions must be drawn appropriately based on the data presented.

Reviewer #1: No

Reviewer #2: Yes

**********

2. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: No

Reviewer #2: Yes

**********

3. Have the authors made all data underlying the findings in their manuscript fully available?

The PLOS Data policy requires authors to make all data underlying the findings described in their manuscript fully available without restriction, with rare exception (please refer to the Data Availability Statement in the manuscript PDF file). The data should be provided as part of the manuscript or its supporting information, or deposited to a public repository. For example, in addition to summary statistics, the data points behind means, medians and variance measures should be available. If there are restrictions on publicly sharing data—e.g. participant privacy or use of data from a third party—those must be specified.

Reviewer #1: No

Reviewer #2: Yes

**********

4. Is the manuscript presented in an intelligible fashion and written in standard English?

PLOS ONE does not copyedit accepted manuscripts, so the language in submitted articles must be clear, correct, and unambiguous. Any typographical or grammatical errors should be corrected at revision, so please note any specific errors here.

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Review Comments to the Author

Please use the space provided to explain your answers to the questions above. You may also include additional comments for the author, including concerns about dual publication, research ethics, or publication ethics. (Please upload your review as an attachment if it exceeds 20,000 characters)

Reviewer #1: This manuscript described NMR measurements of serum samples of diabetic model mice (BKS. Cg db/db) at different age groups with increased severity of atherosclerosis. Although control mice (Jc1:ICR) without diabetes was used to bench mark the carotid arteries and lipid profiles, the NMR measurement of control mice was not included. The experiments appear sound, however, the data processing has clear gaps. Thus, the conclusion is not well founded. I would suggest to address the following concerns before consideration for publication.

1. Need to clarify in details of the principal component analysis (PCA), PC1, PC5, Factor-1, Factor-2 and Factor-3. For example, how to reach to PC1 from FID?

2. Suggest to overlay the NMRs of serum samples of mice to identify fingerprint pattern changes over ages as well as over control mice. Or if signal changes subtracting that of control mice has any relationship with ages?

Reviewer #2: The authors used NMR method to analyze serum samples of diabetic model mice and compared the results with the development of atherosclerosis observed in carotid artery samples. They concluded that revealed that NMR modal properties of serum are associated with arteriosclerotic lesions. Thus, it may be worth exploring the possibility that the risk of cardiovascular events in diabetic patients could be assessed using serum samples.

1. Authors should give more examples for the relation of disease or other pathological condition with NMR modes. What are the pathological mechanisms for these relation? Please mention “Nuclear magnetic resonance (NMR) is a phenomenon that occurs when the nuclei of certain atoms are moved from a static magnetic field to an oscillating magnetic field. Radiofrequency pulse at the resonance frequency causes impulse responses so called free induction decays (FIDs) during NMR measurements. By analyzing FIDs using Fourier transform (FT), we can obtain characteristic spectrums of substances (frequency analysis). Some investigators have studied the utility of this modality in discriminating serum taken from patients with cancers such as renal cell carcinoma, hepatocellular carcinoma and lung cancer” from the reference 18. These knowledge is important for readers to understand the topic.

2. “NMR modal analysis of serum is reported useful for the diagnosis of pancreatic cancer”, please explain how it is useful? Am agree the other reviwers that a direct relation seems not logical. On the other hand there may be indirect relation related to inflammatory markers and hormonal fluctuations associated with the onset and progression of hyperglycemia and carotid atherosclerosis

3. Authors said “We recognize that this study has made it possible to identify NMR mode properties associated with carotid atherosclerosis from the serum of diabetic Mouse models.”. Their answer is not based a scientific perspective. Please give more evidence and explanation that are supported with other publications or their hypothesis. Their hypothesis is very weak to understant this relationship.

Minor points

1. Free induction decay (FID should be indicate in the first use.

**********

6. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: Yes: Bekir Kocazeybek

**********

[NOTE: If reviewer comments were submitted as an attachment file, they will be attached to this email and accessible via the submission site. Please log into your account, locate the manuscript record, and check for the action link "View Attachments". If this link does not appear, there are no attachment files.]

While revising your submission, please upload your figure files to the Preflight Analysis and Conversion Engine (PACE) digital diagnostic tool, https://pacev2.apexcovantage.com/. PACE helps ensure that figures meet PLOS requirements. To use PACE, you must first register as a user. Registration is free. Then, login and navigate to the UPLOAD tab, where you will find detailed instructions on how to use the tool. If you encounter any issues or have any questions when using PACE, please email PLOS at figures@plos.org. Please note that Supporting Information files do not need this step.

Attachment

Submitted filename: comments-plos one.docx

pone.0299641.s005.docx^{(12.4KB, docx)}

PLoS One. 2024 Mar 8;19(3):e0299641. doi: 10.1371/journal.pone.0299641.r003

Author response to Decision Letter 0

26 Jan 2024

1. Need to clarify in details of the principal component analysis (PCA), PC1, PC5, Factor-1, Factor-2 and Factor-3. For example, how to reach to PC1 from FID?

>Response: Thank you for your careful review of our analysis. An explanation of the terms PCA and latent variables (principal components and factors) should facilitate the reader's understanding of the purpose of multivariate analysis in our study. First, we have added an explicit explanation of the significance of PCA in our study to the Methods section.

pp.9, line 195-196. The following sentence has been added: “PCA is a method of dimensionality reduction of data using only explanatory variables and can be used for exploratory analyses such as identifying outliers and potential subgroups [26].”

Next, for the PCA score plots in Figure 3, we were only concerned with explaining PC-1. From our exploratory study, it is possible to explain the group information, although the biological or chemical interpretation of the principal components is difficult. Therefore, we have added an explanation of the group information combining PC-1 and PC-5 to the Results section.

pp.14, line 293-296. The following has been added: “Principal component 1 score was interpreted to be positively associated with week-old progression, whereas principal component 5 score was interpreted to be negatively and moderately associated with week-old progression. Furthermore, on closer examination of principal component 1, it shows …..”

Furthermore, in this study, PLS-DA was used as a two-class classifier. Thus, Factor-1 of the PLS-DA in Figures 4 through 7 is interpreted as an axis that reflects group information; Factor-2 and Factor-3 are significant for maximizing separation on Factor-1 and providing axes that make this visible, but have no specific biological interpretation.

pp.15, line 314-318. The following sentence has been added: (Figure 4B displays …. NMR signal.) “Because PLS-DA was used as a 2-class classifier in the present study, Factor-1 in Figure 4 is interpreted as an axis that reflects group information; axes from Factor-2 onward are interpreted as axes that maximize the separation on Factor-1 and make it visible on the plot. This is also the case for Figures 5 through 7. This interpretation applies to Figures 5 through 7 as well.”

Finally, we consider that the meaningful question "how to reach PC1 from FID?" suggests that readers may have difficulty understanding the analysis process in this paper. For readers unfamiliar with this area, a concise diagram would facilitate understanding of the method. Therefore, we have added a graphical flow chart that focuses primarily on the analysis process in this paper. In this supplemental chart, we have added the articles on PCA, PLS-DA, and STFT to the legend so that readers who wish to review specific methods can do so quickly.

pp.9, line 209-210. Supporting figures and their legends have been added, and the following sentence has been inserted: A graphical flow chart from animal experiments to the multivariate analysis performed in this study was shown as supporting information.

pp.27, line 603-625.

“Supporting information

S1 Fig. Graphical flow chart from the animal experiments to the multivariate analysis

In this figure, the results of NMR analysis of blood samples collected after animal experiments in groups A (n=10) and B (n=10) are shown. In total, 20 1H-NMR FID signal data were acquired, which were short-time Fourier transformed [51], and 20 spectrograms were calculated. For multivariate analysis, a process was performed to obtain a data matrix from the 10 spectrograms. From this data matrix, an exploratory analysis was performed using principal component analysis [26] for the identification of outliers and potential subgroups of animals in groups A and B. From this figure, it was possible to identify subgroups that were not initially anticipated, so PLS-DA [52,53] was performed only with data that corresponded to subgroups in the same data matrix to analyze whether subgroups could be identified. In both principal component analysis and PLS-DA, the variables that were important for the identification of groups in each analysis were plotted as corresponding score plots. This score plot has the same time and frequency axes as the spectrogram and thus can be compared with the spectrogram. The illustration of the mouse and the NMR equipment in this figure was obtained from Research Net, a website where research illustration material is available free of charge (https://www.wdb.com/kenq/illust © WDB Co., Ltd., Tokyo, Japan).

S2 Fig. Graphical flow chart of matrixing of spectrogram data

The process surrounded by the green box in Supporting figure 1 is described in detail. For multivariate analysis of spectrogram data, a 256 × 1024 spectrogram (a) with frequency on the horizontal axis and time on the vertical axis was divided into 1024 single rows (b). All rows were re-arranged into a single row, and 1 × 262,144 single rows were reshaped (c). Since a single row represents spectrogram data for one individual (d), the single rows of spectrogram data for all individuals needed for multivariate analysis are combined to create a data matrix (e).”

>Response: Thank you for your valuable comments. We would like to draw your attention to the score plot that was presented only to the peer reviewers (re-presented below). In this plot, the control and diabetic groups are clearly separated: the variation in the 18-week-old control group is greater than the variation in the 10- to 26-week-old diabetic group, and there are two individuals who are outliers. While this may include an element of strain differences, the results can be interpreted as indicating that the plots of experimental groups with pre- or post-diabetes and atherosclerosis tend to converge more than plots of normal, but with inevitable variation, "control" animals.

[Please see the figure in the accompanying Response to Reviewers file.]]

The identification of signal components in control animals to delineate the boundary between normal and abnormal is one of the most important procedures in bioinformatics research. However, the large variability in the NMR modal properties of the control mice presented in this study makes the process of superposition and subtraction of NMR signals extremely difficult to interpret. Furthermore, age has been noted as one of the intrinsic physiological factors affecting the composition of biological samples from laboratory animals [Bollard ME, Elizabeth G Stanley, John C Lindon, Jeremy K Nicholson, Elaine Holmes. NMR-based metabonomic approaches for evaluating physiological influences on biofluid composition. NMR Biomed. 2005 May;18(3):143-62. PMID: 15627238 DOI: 10.1002/nbm.935]. Thus, the loading plot in Figure 3B can be considered to show NMR modal properties associated with diabetes and atherosclerosis that develop with age.

In conventional spectral analysis, spectral features associated with age or control animal groups may be considered as NMR fingerprint patterns. However, the short-time Fourier transform applied in this study requires arbitrary optimization of the analysis conditions, especially the window function settings, to meet the analysis objectives. Therefore, showing a specific fingerprint pattern from the NMR spectrogram may lead to a great deal of misinterpretation by the reader.

Though the reviewers have provided valuable comments, unfortunately, due to the nature of the STFTs used in this study, we would appreciate your understanding that we are not including the fingerprint pattern changes related to other ages, as well as other control mice.

>Response: Thank you for presenting a topic that contains thought-provoking perspectives in this paper. We would like to express our appreciation. As pointed out, Sato et al. mention that “Some investigators have studied the utility of this modality in discriminating serum taken from patients with cancers such as renal cell carcinoma, hepatocellular carcinoma and lung cancer.”[19] These studies, like our experiments, used proton nuclear magnetic resonance (1H NMR) spectroscopy and human serum samples. On the other hand, unlike our NMR modal analysis method, their studies were NMR metabolomics. In these papers, specific metabolites were identified, offering the prospect that they may be useful for early diagnosis of malignant disease, identification of advanced subgroups, and postoperative surveillance.

Although research on the association between NMR modal analysis methods and diseases or pathological conditions is ongoing, only research on pancreatic cancer has been published as a scientific article at this time [19]. Therefore, unfortunately, we cannot provide specific, biomedical examples of these associations.

Therefore, after careful consideration of the reviewer’s comments, we decided that it was appropriate to explain the pathological mechanisms for these relations from the principles of NMR. The NMR measurements and their analysis in this study are performed on mixtures of low molecular weight materials such as metabolites in biological samples and high molecular weight materials such as serum proteins, without deproteinization in sample preparation or NMR measurements such as T2 filtering. In the method for mixture analysis, Dumez [Ref. 45] points out that “in spectra of mixtures, such as liquid samples or extracts of biological samples, signal overlap prevents reliable identification and accurate quantitation of signals.” His review focuses on homogeneous solutions and on small molecules, which is also a concept in NMR metabolomics. On the other hand, we focus non-selectively on all molecular entities in a biological sample, including protons, which makes the NMR modal analysis method innovative in its ability to comprehensively analyze the information that a biological sample contains.

[Please see the table in the accompanying Response to Reviewers file.]

It is assumed that the NMR mode of biological samples from a group of diseases is affected by biochemical substances (hormones, inflammatory substances, etc.) specific to that disease. As for the pathological mechanism, we believe that, in addition to the abnormal metabolites produced by the substances, the physical properties of macromolecular substances such as albumin, which are affected by the substance profiling produced in this way, also have a significant impact. This attention to this point is a major difference between NMR modal analysis and NMR metabolomics, and in light of the differences with NMR metabolomics, we have added the following text to facilitate understanding of the pathological relevance of NMR modal properties.

pp.19, line 419-427. The following text has been added: “NMR metabolomics experiments focus on low molecular weight substances and obtain their spectra [45]. On the other hand, in NMR modal analysis, all molecular species including protons of up to high molecular weight substances such as serum proteins in addition to low molecular weight metabolites are measured by NMR in a non-selective and comprehensive manner to obtain spectrograms that reflect the state of all hydrogen atoms in the serum sample. This allows us to interpret that biochemical changes due to disease or pathological abnormalities may be reflected in not only static but also dynamic changes of protons in all components in serum (from low to high molecular weight substances and free water). In this study,…..”

>Response: Our intention was to present the NMR modal method as a previous study applied to clinical specimens, but our wording was misleading. As noted, we also considered that “there may be an indirect relation related to inflammatory markers and hormonal fluctuations associated with the onset and progression of hyperglycemia and carotid atherosclerosis”. Therefore, to ensure that our intentions are accurately conveyed, we have revised the text as follows.

pp.4, line 72-76. The following sentence has been added: “As an example, there are published studies in which NMR modal analysis of serum from pancreatic cancer patients yielded results of diagnostic importance [19]. Therefore, we considered the possibility of applying our method not only to pancreatic cancer, but also to other diseases that indirectly or directly alter the physical properties of serum, specifically, the mobility of all hydrogen nuclei contained in serum. In the present study,….”

>Response: As pointed out, we recognize that our statement that this study has made it possible to identify NMR mode associated with carotid atherosclerosis from the serum of diabetic mouse models is somewhat overstated. Considering the results presented by our experiments, our response to this comment is revised based on the results of our experiments as described below: “It is suggested that this study has made it possible to identify NMR mode associated with carotid atherosclerosis from the serum of diabetic mouse models.”

Furthermore, in response, we have also added or revised the text as follows. As pointed out by the reviewers, we have responded from a scientific perspective by adding evidence and explanations supported by other papers and their hypotheses.

Data Analysis Strategy

The purpose of our study was to provide proof-of-concept that the development of atherosclerosis in individuals with a diabetic background can be diagnosed from serum NMR modal experiments. In other words, we consider that identifying serum NMR modes associated with atherosclerotic lesions was key in this experiment. This is essentially different from metabolomics experiments [Alonso-Moreno P, Rodriguez I, Izquierdo-Garcia JL. Benchtop NMR-Based Metabolomics: First Steps for Biomedical Application. Metabolites 2023, 13(5), 614; https://doi.org/10.3390/metabo13050614], which aim to identify biomarkers and understand the underlying metabolic changes associated with various diseases. As in the pancreatic cancer paper, the NMR modal method step in this paper used pattern recognition methods [25]. Since the purpose of this experiment was identification, a combination of PCA and PLS-DA was used.

pp.8-9, line 188-191. The following sentence has been added: “Similar to the study of the diagnosis of pancreatic cancer from serum by applying NMR modal methods [19], pattern recognition [25] of image data was applied to clarify groups from 1H NMR-FID spectrograms of mouse serum in this study, and principal component analysis (PCA) and partial least squares discriminant analysis (PLS-DA) were performed.”

pp.9, line 191-192. The underlined words have been revised as followed: “To perform PCA and PLS-DA, …”

Animal experiments

In order to achieve the goal of identifying atherosclerotic lesions from serum in our experiments, it is crucial that the experiments are carefully planned and well controlled. This is also true for metabolomics experiments, as Ebbels et al. point out [Ref 31], “One of the most important stages, but one which is often overlooked, is the initial design of the experiment.” As reviewer #1 noted, "The experiments appear sound," and we are confident that the animal experiments have been completed successfully enough.

In addition, we performed typical blood biochemistry experiments such as blood glucose, T-cho, TG, etc., as shown in Table 2, but although blood biochemistry results varied from individual to individual, this variation in results did not clearly correlate with the scatter in the PCA score plots in Figure 3.

pp.17, line 368-373. The following sentence has been added: “It can be pointed out that one of the factors contributing to this success was the successful completion of animal experiments conducted under well-controlled conditions, and with the exception of one individual, pathology experiments on mouse carotid arteries were also successful. In fact, it has been pointed out that the first animal experiment is a very important element in the interpretation of the results of multivariate analysis in the step of chemometric studies [31].”

pp.14, line 298-301. The following sentence has been added: “Furthermore, although the blood biochemistry test results shown in Table 2 varied from individual to individual, there is no clear correlation between this variation and the variation in the PCA score plots in Figure 3.”

Principal Component Analysis

If the PCA in Figure 3 formed subgroups by age, the majority of the NMR modes of mouse serum could be interpreted as being influenced by age. However, the 18-week-olds were separated into two groups, and the NMR modes of mouse serum cannot be simply classified by age alone.

To assist in the interpretation of the factors contributing to the separation of the two groups of 18-week-old mice in Figure 3, carotid artery pathology was used in this experiment. Carotid artery pathology was applied to specimens from all ages, not just the 18-week-old individuals, so that lesion evolution could also be confirmed. The results, as shown in Table 1, indicate that 18-week-old mice are in a period of rapid development of carotid atherosclerosis. This result makes us suggest that carotid atherosclerosis may have influenced the distribution of the score plots, although the PCA in Figure 3 may have been affected by aging to a greater or lesser extent.

pp.17, line 374-381. The underlined word and sentences have been added as followed: “Based on the score plots of all PCA data for BKS.Cg db/db mice, all groups (except B18) were divided into a precursor group and progression group by PC-1. In contrast, the data plots for group B18 were scattered and bifurcated into precursor and progression groups, as the atherosclerotic lesions progressed particularly rapidly. If the scatter of the score plots in Figure 3 reflected differences in week age, a distribution by age group would be expected, but the actual results are contrary to this. Since PCA is an analysis that uses only explanatory variables (in this experiment, the serum spectrogram data) for dimensionality reduction [26], it was thought that the results could be interpreted as a visualization of changes in serum status around 18 weeks of age.”

PLS-DA and loading plot

Next, based on the results of carotid artery pathology in 18-week-old mice, the PCA results were divided into several subgroups, and the PLS-DA was used for their classification. The clarity of the separation is shown in Figures 4-7, but it is difficult to determine whether the separation is meaningful or not only from the score plots.

Therefore, a new tool we have prepared for this study is a two-dimensional loading plot based on the spectrogram. This allows us to visualize which areas on the spectrogram are effective pixels for identification. Brereton et al. pointed out the following [Richard G. Brereton, Gavin R. Lloyd.　Partial least squares discriminant analysis: taking the magic away. J.Chemometrics 2014;28(4):213-225. https://doi.org/10.1002/cem.2609]; As an exploratory graphical method that suggests which variables are the most likely to be responsible for discrimination, used often in exploratory studies, PLS-DA has a significant role to play. If the result is that PLS-DA is effective in discriminating in areas such as noise regions where the energy intensity is very weak on the spectrogram, then the results in Figures 4-7 must be interpreted as not particularly meaningful. However, as shown in Figures 4B-7B, the loading plots confirm that the regions correspond to the strong energy regions of the spectrogram as effective regions for identification. This can be interpreted as suggesting that the PLS-DA discrimination in this experiment may be due to differences in serum NMR mode. This step is equivalent to performing a PLS-DA in a metabolomics study to produce a score plot and a corresponding (one-dimensional) loading plot.

pp.17, line 384. The underlined words have been inserted as followed: “…part of these samples was classified into the precursor group. As shown in Fig. 7A, the results for groups B10, B14, and B18 A showed that the samples were divided into these groups by Factor-1.”

pp.17-18, line 387-393. The underlined sentences have been added as followed: “The PLS-DA score plot for this precursor group shown in Figure 7A cannot be explained by age-related changes alone. In this study, time-trend changes in carotid artery pathology were observed, which, as noted above, showed rapid atherosclerotic lesion progression at 18 weeks. Considering the characteristics of the distribution of the B18 group shown on some score plots, we interpret this as suggesting that the clustering and separation of samples shown in the multivariate score plots in this study may be related to the progression of atherosclerotic lesions, though there is a certain degree of influence of differences in age.”

pp.18, line 394-397. The underlined sentences have been added and revised as followed: “In addition, the time-frequency domains of the variables valid for discrimination shown in the loading plots in Figures 3-7 corresponded roughly to the signal domains in the serum spectrograms in Figure 2. These results suggest that the progression of arteriosclerotic lesions was reflected in the serum NMR modal properties.”

Interpretation of NMR mode

The purpose of this experiment was different from that of metabolomic analysis, so we will not identify the metabolite responsible for the identification. On the other hand, we conducted carotid artery pathology examinations in mice and interpreted the NMR mode based on the histopathology findings.

The changes in carotid artery pathology at and around 18 weeks of age naturally lead us to infer a pathophysiological mechanism by which the properties of mouse serum would be affected by inflammatory changes and hormonal fluctuations during this period. We interpreted this change in blood properties, either directly or indirectly, as a major variation in serum NMR mode at 18 weeks of age. Because the present experimental data, unlike metabolomic analysis, are of interest for all molecular species (low to high molecular weight substances and free water), including protons in serum, and can be regarded as reflecting molecular species and their amounts, as well as intra- and intermolecular forces acting on the protons in each molecule, the NMR modal analysis method has the potential to sensitively observe differences in the physical properties of the serum as a whole. Based on appropriately controlled animal studies, we hypothesize that the differences in NMR modal properties shown in this study are due to mouse carotid atherosclerosis.

Since the content is related to reviewer #2's comment 1, the addition of the manuscript is the same as in “pp.19, line 419-427” above.

pp.19, line 419-427. The following sentence has been added: “NMR metabolomics experiments focus on low molecular weight substances and obtain their spectra. On the other hand, in NMR modal analysis, all molecular species including protons of up to high molecular weight substances such as serum proteins in addition to low molecular weight metabolites are measured by NMR in a non-selective and comprehensive manner to obtain spectrograms that reflect the state of all hydrogen atoms in the serum sample. This allows us to interpret that biochemical changes due to disease or pathological abnormalities may be reflected in not only static but also dynamic changes of all protons in components in serum (from low to high molecular weight substances and free water).”

Issues in the future

According to Bollard et al, "Each of the spectral integrals can be regarded as a metabolic descriptor which can be used to investigate similarities or differences in the data, according to an individual's biochemistry.” Namely, in metabolomics research, spectrum-based loading plots and NMR spectral fingerprints can be matched to interpret metabolic changes associated with various diseases.

On the other hand, the physicochemical meaning of the calculations of the NMR spectrograms will be clarified in further studies. The next question is what kind of biomedical interpretation of spectrogram-based loading plots will be possible once the meaning of the spectrograms is clarified. Furthermore, there are several issues to be addressed, including room for improvement in statistical methods, larger sample size for model building, the relevance of NMR mode properties to diseases and pathologies, and the applicability of NMR to more complex human clinical specimens. However, despite some of the challenges, the current paper is positioned as a feasibility study for future serum NMR modal analysis and diagnosis of carotid artery extension in more complex human blood specimens.

The above response is meant as an explanation to the reviewers and does not result in revisions or additions to the manuscript.

Minor points

1. Free induction decay (FID should be indicate in the first use.

pp.7, line 142. The underlined words have been revision where FID was used for the first time: “Free induction decay (FID) data collection for NMR measurement…”

References

For changes in reference numbers made to accommodate the addition of references, please see the table in the accompanying Response to Reviewers file.

Attachment

Submitted filename: ResponseToReviewers.docx

pone.0299641.s006.docx^{(118.5KB, docx)}

PLoS One. doi: 10.1371/journal.pone.0299641.r004

Decision Letter 1

Jorddy Neves Cruz

13 Feb 2024

Time-frequency analysis reveals an association between the specific nuclear magnetic resonance (NMR) signal properties of serum samples and arteriosclerotic lesion progression in a diabetes mouse model

PONE-D-23-28528R1

Dear Dr. Kanawaku,

We’re pleased to inform you that your manuscript has been judged scientifically suitable for publication and will be formally accepted for publication once it meets all outstanding technical requirements.

Within one week, you’ll receive an e-mail detailing the required amendments. When these have been addressed, you’ll receive a formal acceptance letter and your manuscript will be scheduled for publication.

An invoice for payment will follow shortly after the formal acceptance. To ensure an efficient process, please log into Editorial Manager at http://www.editorialmanager.com/pone/, click the 'Update My Information' link at the top of the page, and double check that your user information is up-to-date. If you have any billing related questions, please contact our Author Billing department directly at authorbilling@plos.org.

If your institution or institutions have a press office, please notify them about your upcoming paper to help maximize its impact. If they’ll be preparing press materials, please inform our press team as soon as possible -- no later than 48 hours after receiving the formal acceptance. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

Kind regards,

Jorddy Neves Cruz

Academic Editor

PLOS ONE

Additional Editor Comments (optional):

Reviewers' comments:

Reviewer's Responses to Questions

Comments to the Author

1. If the authors have adequately addressed your comments raised in a previous round of review and you feel that this manuscript is now acceptable for publication, you may indicate that here to bypass the “Comments to the Author” section, enter your conflict of interest statement in the “Confidential to Editor” section, and submit your "Accept" recommendation.

Reviewer #1: All comments have been addressed

Reviewer #2: All comments have been addressed

**********

2. Is the manuscript technically sound, and do the data support the conclusions?

Reviewer #1: Yes

Reviewer #2: Yes

**********

3. Has the statistical analysis been performed appropriately and rigorously?

Reviewer #1: N/A

Reviewer #2: Yes

**********

4. Have the authors made all data underlying the findings in their manuscript fully available?

Reviewer #1: Yes

Reviewer #2: Yes

**********

5. Is the manuscript presented in an intelligible fashion and written in standard English?

Reviewer #1: Yes

Reviewer #2: Yes

**********

6. Review Comments to the Author

Reviewer #1: The patern learning is an interesting way to explain the data but the direct molecular bases need to be established before it can be useful. Encourage to do further studies and communicate later as a followup.

Reviewer #2: All revisions were performed and the manuscript can be understandabl now, in this current form.T he authors also respond my comments in a scientific perspective.

**********

7. PLOS authors have the option to publish the peer review history of their article (what does this mean?). If published, this will include your full peer review and any attached files.

If you choose “no”, your identity will remain anonymous but your review may still be made public.

Do you want your identity to be public for this peer review? For information about this choice, including consent withdrawal, please see our Privacy Policy.

Reviewer #1: No

Reviewer #2: Yes: Bekr Kocazeybek

**********

PLoS One. doi: 10.1371/journal.pone.0299641.r005

Acceptance letter

Jorddy Neves Cruz

26 Feb 2024

PONE-D-23-28528R1

PLOS ONE

Dear Dr. Kanawaku,

I'm pleased to inform you that your manuscript has been deemed suitable for publication in PLOS ONE. Congratulations! Your manuscript is now being handed over to our production team.

At this stage, our production department will prepare your paper for publication. This includes ensuring the following:

* All references, tables, and figures are properly cited

* All relevant supporting information is included in the manuscript submission,

* There are no issues that prevent the paper from being properly typeset

If revisions are needed, the production department will contact you directly to resolve them. If no revisions are needed, you will receive an email when the publication date has been set. At this time, we do not offer pre-publication proofs to authors during production of the accepted work. Please keep in mind that we are working through a large volume of accepted articles, so please give us a few weeks to review your paper and let you know the next and final steps.

Lastly, if your institution or institutions have a press office, please let them know about your upcoming paper now to help maximize its impact. If they'll be preparing press materials, please inform our press team within the next 48 hours. Your manuscript will remain under strict press embargo until 2 pm Eastern Time on the date of publication. For more information, please contact onepress@plos.org.

If we can help with anything else, please email us at customercare@plos.org.

Thank you for submitting your work to PLOS ONE and supporting open access.

Kind regards,

PLOS ONE Editorial Office Staff

on behalf of

Dr. Jorddy Neves Cruz

Academic Editor

PLOS ONE

Associated Data

This section collects any data citations, data availability statements, or supplementary materials included in this article.

Supplementary Materials

S1 Fig. Graphical flow chart from the animal experiments to the multivariate analysis.

References (cited only in the Supporting Information)

51. Wacker M, Witte H. Time-frequency techniques in biomedical signal analysis. a tutorial review of similarities and differences. Methods Inf Med. 2013;52:279–296.

52. Barker M, Rayens W. Partial least squares for discrimination. J Chemo. 2003;17:166–173.

53. Sylvie Chevallier S, Bertrand D, Kohler A, Courcoux P. Application of PLS-DA in multivariate image analysis. J Chemo. 2006;20: 221–229.

(TIF)

pone.0299641.s001.tif^{(464KB, tif)}

S2 Fig. Graphical flow chart of matrixing of spectrogram data.

(TIF)

pone.0299641.s002.tif^{(563.4KB, tif)}

S1 Data

(ZIP)

pone.0299641.s003.zip^{(7.7MB, zip)}

Attachment

Submitted filename: ResponseToReviewers.docx

pone.0299641.s004.docx^{(119.1KB, docx)}

Attachment

Submitted filename: comments-plos one.docx

pone.0299641.s005.docx^{(12.4KB, docx)}

Attachment

Submitted filename: ResponseToReviewers.docx

pone.0299641.s006.docx^{(118.5KB, docx)}

Data Availability Statement

All relevant data are within the paper and its Supporting information files.

[pone.0299641.ref001] 1.Grundy SM, Benjamin IJ, Burke GL, Chait A, Eckel RH, Howard BV, et al. Diabetes and cardiovascular disease: a statement for healthcare professionals from the American Heart Association. Circulation. 1999;100: 1134–1146. doi: 10.1161/01.cir.100.10.1134 [DOI] [PubMed] [Google Scholar]

[pone.0299641.ref002] 2.Gæde P, Lund-Andersen H, Parving H-H, Pedersen O. Effect of a multifactorial intervention on mortality in type 2 diabetes. N Engl J Med. 2008;358: 580–591. doi: 10.1056/NEJMoa0706245 [DOI] [PubMed] [Google Scholar]

[pone.0299641.ref003] 3.Zureik M, Bureau JM, Temmar M, Adamopoulos C, Courbon D, Bean K, et al. Echogenic carotid plaques are associated with aortic arterial stiffness in subjects with subclinical carotid atherosclerosis. Hypertension. 2004;41: 519–527. [DOI] [PubMed] [Google Scholar]

[pone.0299641.ref004] 4.Antiga L, Wasserman BA, Steinman DA. On the overestimation of early wall thickening at the carotid bulb by black blood MRI, with implications for coronary and vulnerable plaque imaging. Magn Reson Med. 2008;60: 1020–1028. doi: 10.1002/mrm.21758 [DOI] [PubMed] [Google Scholar]

[pone.0299641.ref005] 5.Bartlett ES, Walters TD, Symons SP, Fox AJ. Quantification of carotid stenosis on CT angiography. Am J Neuroradiol. 2006;27: 13–19. [PMC free article] [PubMed] [Google Scholar]

[pone.0299641.ref006] 6.Ueda M, Takabe M, Sawada T, Urai S, Hashimoto N, Ohara T. Comparative study of vascular endothelial function tests (FMD, RHI) in patients with type 2 diabetes. J Diabetes. 2017;60: 370–377. [Google Scholar]

[pone.0299641.ref007] 7.Fukui S. Tests to be performed for prevention of arteriosclerosis in physical health checkups. Ningen Dokku. 2016;30: 809–821. [Google Scholar]

[pone.0299641.ref008] 8.The Japan Diabetes Society, Guidelines for Diabetes Care 2019. Nankkodo. 2019: 201–218.

[pone.0299641.ref009] 9.The Japanese Society of Hypertension, Committee for the Preparation of Guidelines for the Treatment of Hypertension, Guidelines for the Treatment of Hypertension 2019, Japanese Society of Hypertension, 2019: 25–30.

[pone.0299641.ref010] 10.Okada E, Oida K, Tada H, Asazuma K, Eguchi K, Tohda G, et al. Hyperhomocysteinemia is a risk factor for coronary arteriosclerosis in Japanese patients with type 2 diabetes. Diabetes Care. 1999;22: 484–490. doi: 10.2337/diacare.22.3.484 [DOI] [PubMed] [Google Scholar]

[pone.0299641.ref011] 11.Lopes-Virella MF, Virella G. Cytokines, modified lipoproteins, and arteriosclerosis in diabetes. Diabetes. 1996;45: 40–45. doi: 10.2337/diab.45.3.s40 [DOI] [PubMed] [Google Scholar]

[pone.0299641.ref012] 12.Lundberg AM, Hansson GK. Innate immune signals in atherosclerosis. Clin Immunol. 2010;134: 5–24. doi: 10.1016/j.clim.2009.07.016 [DOI] [PubMed] [Google Scholar]

[pone.0299641.ref013] 13.López-Tinoco C, Roca M, Fernández-Deudero A, García-Valero A, Bugatto F, Aguilar-Diosdado M, et al. Cytokine profile, metabolic syndrome and cardiovascular disease risk in women with late-onset gestational diabetes mellitus. Cytokine. 2021;58: 14–19. [DOI] [PubMed] [Google Scholar]

[pone.0299641.ref014] 14.Severino P, D’Amato A, Netti L, Pucci M, Infusino F, Maestrini V, et al. Myocardial ischemia and diabetes mellitus: role of oxidative stress in the connection between cardiac metabolism and coronary blood flow. J Diabetes Res. 2019;2019: 9489826. doi: 10.1155/2019/9489826 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0299641.ref015] 15.Liguori A, Abete P, Hayden JM, Cacciatore F, Rengo F, Ambrosio G, et al. Effect of glycaemic control and age on low-density lipoprotein susceptibility to oxidation in diabetes mellitus type 1. Eur Heart J. 2001;22: 2075–2084. doi: 10.1053/euhj.2001.2655 [DOI] [PubMed] [Google Scholar]

[pone.0299641.ref016] 16.Bierhaus A, Hofmann MA, Ziegler R, Nawroth PP. AGEs and their interaction with AGE-receptors in vascular disease and diabetes mellitus. I. The AGE concept. Cardiovasc Res. 1998;37: 586–600. doi: 10.1016/s0008-6363(97)00233-2 [DOI] [PubMed] [Google Scholar]

[pone.0299641.ref017] 17.Singh RB, Mengi SA, Xu Y-J, Arneja AS, Dhalla NS. Pathogenesis of atherosclerosis: A multifactorial process. Exp Clin Cardiol. 2002;7: 40. [PMC free article] [PubMed] [Google Scholar]

[pone.0299641.ref018] 18.Hirakawa K, Koike K, Kanawaku Y, Moriyama T, Sato N, Suzuki T, et al. Short-time Fourier transform of free induction decays for the analysis of serum using proton nuclear magnetic resonance. J Oleo Sci. 2019;68: 369–378. doi: 10.5650/jos.ess18212 [DOI] [PubMed] [Google Scholar]

[pone.0299641.ref019] 19.Sato A, Masui T, Yogo A, Ito T, Hirakawa K, Kanawaku Y, et al. Time-frequency analysis of serum with proton nuclear magnetic resonance for diagnosis of pancreatic cancer. Sci Rep. 2020;10: 21941. doi: 10.1038/s41598-020-79087-3 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0299641.ref020] 20.Zimmet P, Alberti KGMM, Shaw J. Global and societal implications of the diabetes epidemic. Nature. 2001;414: 782–787. doi: 10.1038/414782a [DOI] [PubMed] [Google Scholar]

[pone.0299641.ref021] 21.Ray NF, Thamer M, Gardner E, Chan JK. Economic consequences of diabetes mellitus in the U.S. in 1997. American Diabetes Association. Diabetes Care. 1998;21: 296–309. [DOI] [PubMed] [Google Scholar]

[pone.0299641.ref022] 22.Zheng H, Zheng Y, Zhao L, Chen M, Bai G, Hu Y, et al. Cognitive decline in type 2 diabetic db/db mice may be associated with brain region-specific metabolic disorders. Biochim Biophys Acta Mol Basis Dis. 2017;1863: 266–273. doi: 10.1016/j.bbadis.2016.11.003 [DOI] [PubMed] [Google Scholar]

[pone.0299641.ref023] 23.Tajima T, Hirakawa K, Kawaguchi H, Sakamoto A. Proton nuclear magnetic resonance and pattern recognition analysis of liver extracts from rats under different anesthetics. BMC Med Imaging. 2012;12: 28. doi: 10.1186/1471-2342-12-28 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0299641.ref024] 24.Short-Time Fourier Transform, The MathWorks, Inc. [cited 24 Aug 2023]. https://jp.mathworks.com/help/signal/ref/stft.html?lang=en

[pone.0299641.ref025] 25.Amigo JM, Babamoradi H, Elcoroaristizabal S. Hyperspectral image analysis. A tutorial. Anal Chim Acta. 2015:896:34–51. doi: 10.1016/j.aca.2015.09.030 [DOI] [PubMed] [Google Scholar]

[pone.0299641.ref026] 26.Wold S, Esbensen K, Geladi P. Principal component analysis. Chemometr Intell Lab Syst. 1987; 2:37–52. [Google Scholar]

[pone.0299641.ref027] 27.Kawaguchi H, Hirakawa K, Miyauchi K, Koike K, Ohno Y, Sakamoto A. Pattern recognition analysis of proton nuclear magnetic resonance spectra of brain tissue extracts from rats anesthetized with propofol or isoflurane. PLoS One. 2010;5: e11172. doi: 10.1371/journal.pone.0011172 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0299641.ref028] 28.Goodpaster AM, Kennedy MA. Quantification and statistical significance analysis of group separation in NMR-based metabolomics studies. Chemometr Intell Lab Syst. 2011;109: 162–170. [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0299641.ref029] 29.Dickens AM, Larkin JR, Davis BG, Griffin JL, Claridge TD, Sibson NR, et al. NMR-based metabolomics separates the distinct stages of disease in a chronic relapsing model of multiple sclerosis. J Neuroimmune Pharmacol. 2015;10: 435–444. doi: 10.1007/s11481-015-9622-0 [DOI] [PubMed] [Google Scholar]

[pone.0299641.ref030] 30.Asano T, Hirakawa K, Koike K, Ohno Y, Fujino O. Visualization of different characteristics of cerebrospinal fluid with acute encephalopathy and febrile seizures using pattern recognition analysis of 1H NMR. Pediatr Res. 2015;77: 70–74. doi: 10.1038/pr.2014.141 [DOI] [PubMed] [Google Scholar]

[pone.0299641.ref031] 31.Ebbels TMD, Cavill R. Bioinformatic methods in NMR-based metabolic profiling. Prog Nucl Magn Reson Spectrosc. 2009;55:361–374. [Google Scholar]

[pone.0299641.ref032] 32.de Graaf RA. In Vivo NMR Spectroscopy: Principles and Techniques. Chichester, England. John Wiley & Sons; 2018. pp. 15–17. [Google Scholar]

[pone.0299641.ref033] 33.Jung JY, Lee HS, Kang DG, Kim NS, Cha MH, Bang OS, et al. 1H-NMR-based metabolomics study of cerebral infarction. Stroke. 2011;42: 1282–1288. doi: 10.1161/STROKEAHA.110.598789 [DOI] [PubMed] [Google Scholar]

[pone.0299641.ref034] 34.Okuda T, Hirakawa K, Hakan O, Ohno Y, Hyakusoku H. Nuclear magnetic resonance spectroscopy as a useful tool for ingredient analysis of synthetic materials. Plast Reconstr Surg. 2012;129: 197e–198e. doi: 10.1097/PRS.0b013e3182365d00 [DOI] [PubMed] [Google Scholar]

[pone.0299641.ref035] 35.Sinclair AJ, Viant MR, Ball AK, Burdon MA, Walker EA, Stewart PM, et al. NMR-based metabolomic analysis of cerebrospinal fluid and serum in neurological diseases—a diagnostic tool? NMR Biomed. 2010;23: 123–132. doi: 10.1002/nbm.1428 [DOI] [PubMed] [Google Scholar]

[pone.0299641.ref036] 36.Mora-Ortiz M, Nunez Ramos P, Oregioni A, Claus SP. NMR metabolomics identifies over 60 biomarkers associated with type II diabetes impairment in db/db mice. Metabolomics. 2019;15: 89. doi: 10.1007/s11306-019-1548-8 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0299641.ref037] 37.Khattri RB, Thome T, Ryan TE. Tissue-specific (1)H-NMR metabolomic profiling in mice with adenine-induced chronic kidney disease. Metabolites. 2021;11: 45. doi: 10.3390/metabo11010045 [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0299641.ref038] 38.Graham SF, Holscher C, McClean P, Elliott CT, Green BD. 1H NMR metabolomics investigation of an Alzheimer’s disease (AD) mouse model pinpoints important biochemical disturbances in brain and plasma. Metabolomics. 2013;9: 974–983. [Google Scholar]

[pone.0299641.ref039] 39.Emwas A-HM, Salek RM, Griffin JL, Merzaban J. NMR-based metabolomics in human disease diagnosis: applications, limitations, and recommendations. Metabolomics. 2013;9: 1048–1072. [Google Scholar]

[pone.0299641.ref040] 40.Kromke M, Palomino-Schätzlein M, Mayer H, Pfeffer S, Pineda-Lucena A, Luy B, et al. Profiling human blood serum metabolites by nuclear magnetic resonance spectroscopy: a comprehensive tool for the evaluation of hemodialysis efficiency. Transl Res. 2016;171: 71–82.e71-79. doi: 10.1016/j.trsl.2016.02.003 [DOI] [PubMed] [Google Scholar]

[pone.0299641.ref041] 41.Verwaest KA, Vu TN, Laukens K, Clemens LE, Nguyen HP, Van Gasse B, et al. (1)H NMR based metabolomics of CSF and blood serum: a metabolic profile for a transgenic rat model of Huntington disease. Biochim Biophys Acta. 2011;1812: 1371–1379. doi: 10.1016/j.bbadis.2011.08.001 [DOI] [PubMed] [Google Scholar]

[pone.0299641.ref042] 42.Hatazawa J. Clinical application of proton magnetic resonance spectroscopy in neurological diseases. AJNR Am J Neuroradiol. 1996;17: 737–747.8730195 [Google Scholar]

[pone.0299641.ref043] 43.Yang Y, Nagarajaiah S. Time-frequency blind source separation using independent component analysis for output-only modal identification of highly damped structures. J Struct Eng. 2013;139: 1780–1793. [Google Scholar]

[pone.0299641.ref044] 44.Perez-Ramirez CA, Amezquita-Sanchez JP, Adeli H, Valtierra-Rodriguez M, Romero-Troncoso RdJ, Dominguez-Gonzalez A, et al. Time-frequency techniques for modal parameters identification of civil structures from acquired dynamic signals. J Vibroengineering. 2016;18: 3164–3185. [Google Scholar]

[pone.0299641.ref045] 45.Dumez JN. NMR methods for the analysis of mixtures. Chem Commun (Camb). 2022; 58:13855–13872. doi: 10.1039/d2cc05053f [DOI] [PMC free article] [PubMed] [Google Scholar]

[pone.0299641.ref046] 46.Beckman JA, Creager MA, Libby P. Diabetes and atherosclerosis. J Am Med Assoc. 2002;297: 2570–2581. [DOI] [PubMed] [Google Scholar]

[pone.0299641.ref047] 47.Jiang G, Luk AOY, Tam CHT, Xie F, Carstensen B, Lau ESH, et al. Progression of diabetic kidney disease and trajectory of kidney function decline in Chinese patients with type 2 diabetes. Kidney Int. 2019;95: 178–187. doi: 10.1016/j.kint.2018.08.026 [DOI] [PubMed] [Google Scholar]

[pone.0299641.ref048] 48.Yagihashi S, Terauchi Y, Mori Y, Ikegami H, Satoh Y. Diagnostic criteria of diabetes in laboratory animals. Journal of the Japan Diabetes Society. 2010;53: 379–384 (Japanese). [cited 24 Aug 2023]. Available from: https://www.jstage.jst.go.jp/article/tonyobyo/53/5/53_5_379/_pdf [Google Scholar]

[pone.0299641.ref049] 49.Shinohara M, Masada T, Oikawa T, Masuyama T, Takahashi T, Katsuta Y, et al. Diabetes mellitus in a female SDT rat model of spontaneous type 2 diabetes mellitus. Diabetes. 2004;47: 111–116. [Google Scholar]

[pone.0299641.ref050] 50.2003 Laboratory Animal Data Collection, Japan SLC, Inc. (Shizuoka, Japan) [cited 24 Aug 2023]. http://www.jslc.co.jp/pdf/mouse/019_akita.pdf.

PERMALINK

Time-frequency analysis reveals an association between the specific nuclear magnetic resonance (NMR) signal properties of serum samples and arteriosclerotic lesion progression in a diabetes mouse model

Kanako Yui

Yoshimasa Kanawaku

Akio Morita

Keiko Hirakawa

Fanlai Cui

Roles

Abstract

Introduction

Background

Objective

Method

Breeding of experimental animals

Understanding of blood glucose dynamics

Weeks of sampling and sample size

Sample collection and tests performed

Pathological testing

Free induction decay data collection for NMR measurement

NMR data analysis

Multivariate analysis

Results

Pathological results

Table 1. Summary of Jcl:ICR and BKS.Cg db/db carotid artery lesion evaluation.

Fig 1. H&E staining of carotid artery samples.

Blood biochemistry tests

Table 2. Results of biochemical tests performed on two BKS.Cg db/db mice at each week of age.

NMR modal analysis

Fig 2. Representative spectrogram of an 18-week-old BKS.Cg db/db mouse (B18-1).

Fig 3. Results of principal component analysis using spectrogram data from all BKS.Cg db/db serum samples.

Fig 4. Results of partial least squares discriminant analysis performed on spectrogram data for serum samples from 10- and 14-week-old and 22- and 26-week-old mice that grouped as shown in Fig 3A.

Fig 5. Results of partial least squares discriminant analysis performed on spectrogram data for serum samples from mice at 18 weeks of age only.

Fig 7. Results of partial least squares discriminant analysis performed on spectrogram data of serum samples from 10-, 14-, and 18-week-old mice.

Fig 6. Results of partial least squares discriminant analysis performed on spectrogram data for all serum samples of BKS.Cg db/db mice.

Discussion

Conclusions

Supporting information

Acknowledgments

Data Availability

Funding Statement

References

Author response to previous submission

Decision Letter 0

Jorddy Neves Cruz

Roles

Author response to Decision Letter 0

Decision Letter 1

Jorddy Neves Cruz

Roles

Acceptance letter

Jorddy Neves Cruz

Roles

Associated Data

Supplementary Materials

Data Availability Statement

ACTIONS

PERMALINK

RESOURCES

Similar articles

Cited by other articles

Links to NCBI Databases