Table 3.

Summary of metabolomic techniques and examples of their applications in cancer research

Technique	Strengths	Limitations	Related applications in cancer research	Information of samples	Result and significance in PPPM
NMR	Nondestructively analyze samples either in body fluids or in vivo	Low sensitivity	Madhu et al. 2016 [202]	Ten benign prostate tissue samples, seven prostate cancer (PCa) specimens from untreated patients, six PCa specimens from patients treated with Degarelix	This study demonstrated the concentration of specific metabolites could reflect the real-time response of antitumor drug treatment
	High reproducibility and repeatability	Poor quantification ability	Hajduk et al. 2016 [203]	Blood sample form 45 head and neck squamous cell carcinoma patients with radiotherapy (RT) or chemoradiotherapy (CHRT)	This study monitoring the effect of RT based on metabolomics method provide the basis of precision treatment
	Quantification analysis of metabolites	Requires large sample size
GC-MS	Especially suitable for thermostable and volatile and nonpolar metabolites	Derivatization required, so unfit for polar metabolites such as polyphenos and glycosides
	High separation efficiency and reproducibility	Extensive sample preparation steps and time consuming	Hadi et al. 2017 [204]	Serum sample from 152 pre-operative breast cancer (BC) patients and 155 healthy controls	This study constructed models using distinct metabolites to diagnose, stage, grade and evaluate neoadjuvant status providing metabolic evidence for early diagnosis and treatment of BC
	Very sensitive	Destructive (sample not recoverable)	Cameron et al. 2016 [205]	Sputum sample from 34 suspected lung cancer (LC) patients, 33 healthy controls	This study demonstrated the feasibility of sputum metabolomics analysis and indicated this method could help ones to noninvasively screen the high-risk population of lung cancer
	High mass accuracy to detect compounds	Derived samples can only be stored for 2-3 days
	Highly developed compound libraries and software for metabolite identification	Novel compound identification is difficult
	Can be mostly automated	Cannot be used in imaging
LC-MS	Be capable to detect the largest potion of metabolome	Lower separation power and reproducibility than GC-MS	Di Gangi et al. 2016 [206]	Serum sample from 40 suspected pancreatic cancer patients and 40 healthy controls	This research identified several metabolites as highly discriminative potential prognostic markers
	Excellent sensitivity	Destructive to samples	Hou et al. 2014 [207]	Plasma from 38 cervical cancer patients with different response to neoadjuvant chemotherapy (NACT)	A prediction model with an AUC of 0.9407 can be used to predict the patient’s response to NACT, which has important implications in personalized treatment and outcomes
	Simple sample preparation and short separation time	Not very been quantified	Mathé et al. 2014 [208]	Urine collected from 469 patients with lung cancer and 536 population controls	Creatine riboside and N-acetylneuraminic acid can be regarded as novel noninvasive biomarkers for the early diagnosis and prognosis of lung cancer
	Detects a wider range of metabolites than GC-MS	High instrumental cost
	Analysis of more polar compounds without derivatization and ideal for nonvolatile compounds	More instrumental variables than in NMR and GC-MS