GWA-based pleiotropic analysis identified potential SNPs and genes related to type 2 diabetes and obesity

Abstract

Metabolic syndrome is a cluster of symptoms including excessive body fat and insulin resistance which may lead to obesity and type 2 diabetes (T2D). The physiological and pathological cross-talk between T2D and obesity is crucial and complex, meanwhile, the genetic connection between T2D and obesity is largely unknown. The purpose of this study is to identify pleiotropic SNPs and genes between these two associated conditions by applying genetic analysis incorporating pleiotropy and annotation (GPA) on two large genome-wide association studies (GWAS) data sets: a body mass index (BMI) data set containing 339,224 subjects and a T2D data set containing 110,452 subjects. In all, 5182 SNPs showed pleiotropy in both T2D and obesity. After further prioritization based on suggested local false discovery rates (FDR) by the GPA model, 2146 SNPs corresponding to 217 unique genes are significantly associated with both traits (FDR < 0.2), among which 187 are newly identified pleiotropic genes compare with original GWAS in individual traits. Subsequently, gene enrichment and pathway analyses highlighted several pleiotropic SNPs including rs849135 (FDR = 0.0002), rs2119812 (FDR = 0.0018), rs4506565 (FDR = 1.23E–08), rs1558902 (7.23E–10) and corresponding genes JAZF1, SYN2, TCF7L2, FTO which may play crucial rol5es in the etiology of both T2D and obesity. Additional evidences from expression data analysis of pleiotropic genes strongly supports that the pleiotropic genes including JAZF1 (p = 1.39E–05 and p = 2.13E–05), SYN2 (p = 5.49E–03 and p = 5.27E–04), CDKN2C (p = 1.99E–12 and p = 6.27E–11), RABGAP1 (p = 3.08E–03 and p = 7.46E–03), and UBE2E2 (p = 1.83E–04 and p = 8.22E–03) play crucial roles in both obesity and T2D pathogenesis. Pleiotropic analysis integrated with functional network identified several novel and causal SNPs and genes involved in both BMI and T2D which may be ignored in single GWAS.

Introduction

Metabolic syndrome (MetS) is a cluster of symptoms including excessive body fat, insulin resistance, glucose intolerance, increased blood pressure, hyperlipidemia, and hyperglycemia, all of which increase the risk of serious diseases such as central obesity and type 2 diabetes (T2D) [1,2]. Over the past two decades, the number of people with some degree of MetS has increased tremendously [3]. Although the worldwide prevalence of MetS ranges from 10% to up to 84% due to different hereditary backgrounds and living environments [4, 5], the International Diabetes Federation evaluates that a quarter of adult population in the world are suffering MetS [3].

MetS-associated diseases cause a high rate of morbidity and mortality in a considerable portion of the population. For instance, high blood pressure was supported to be the number one risk factor which accounts for 9.4 million mortalities and 7% morbidity worldwide in 2010 [6]. Both obesity and T2D could increase the risk of cardiovascular morbidity and mortality [7, 8]. As the most prevalent symptoms of MetS, obesity and T2D are becoming serious health problems mainly characterized by excessive body fat and insulin resistance, respectively. The body mass index (BMI), defined as weight divided by height squared (kg/m²), is used to qualify the amount of fat tissue mass and is widely used in both research and clinical settings regarding obesity. T2D is a long-term metabolic disorder characterized mainly by high blood sugar, insulin resistance, and relative lack of insulin. Both environmental and genetic factors play significant roles in the development of obesity and T2D [9]. Changes in the global economy have altered the average person’s lifestyle in both diet and physical activities, which in combination with hereditary factors have led to a worldwide increase in obesity and T2D [1]. The global prevalence of obesity has nearly doubled from 1980 to 2008 [10]. Meanwhile, until 2010, an estimated 285 million people worldwide had diabetes mellitus, the number may reach up to 439 million by 2030 [11], 90% of whom had T2D [12]. The physiological and pathological cross-talk between T2D and obesity is complex; for example, obesity is a potential inducer of numerous serious diseases including T2D, but T2D may also cause unexplained weight loss [13].

Genome-wide association studies (GWAS) have been performed in T2D and obesity individually, which identified a number of SNPs/loci contributing to these two traits and provided a large amount of GWAS data [14, 15]. However, the specific genetic connection between T2D and obesity is largely unknown. Accumulating evidence from GWAS suggests that different complex diseases often share common genetic risk bases [16]. Pleiotropy is defined as one gene influencing two or more phenotypic traits. A mutation in a pleiotropic gene may influence two or more traits simultaneously [17, 18]. In this study, we applied a recently developed novel genetic analysis incorporating pleiotropy and annotation (GPA) model on two large GWAS data sets for pleiotropic analysis for obesity and T2D. GPA can increase statistical power to identify common risk variants through joint analysis of summary statistics of multiple GWAS data sets for different traits [16].

Gene-based pathway and network analysis can provide more topological information between genes and corresponding traits, which allow us to better understand the physiology and pathology of T2D and obesity. Pathways/modules generated by network analysis give us a comprehensive view between causal genes and biological process [19]. There are several bioinformatics tools such as Database for Annotation, Visualization and Integrated Discovery (DAVID), Search Tool for the Retrieval of Interacting Genes/Proteins (STRING), and Cytoscape can be selected for multiple function of pathway and network analysis [20, 21]. Specific description of the above bioinformatics tools can be find in our previous publication [22, 23].

In this study, GWAS-based pleiotropic analysis, integrated with gene enrichment, and pathway analysis shed light on the genetic link between T2D and obesity. In order to gain additional and functional mechanistic support, we performed expression data analysis of pleiotropic genes.

Materials and methods

GWAS data sets

The summary statistical results of two GWAS data sets were used in this study. The T2D GWAS data set was generated from a trans-ethnic meta-analysis in 110,452 subjects from four ancestries including 12,171 cases and 56,862 controls of European ancestry; 6952 cases and 11,865 controls of East Asian ancestry; 5561 cases and 14,458 controls of South Asian ancestry; and 1804 cases and 779 controls of Mexican and Mexican American ancestry (a total of 26,488 cases and 83,964 controls) [15], which contains a total number of about 2.9 million SNPs. The data set was downloaded from DIAbetes Genetics Replication and Meta-analysis Consortium (http://diagram-consortium.org/downloads.html). The BMI data set was generated from meta-analysis of 82 GWAS and 43 metabochip studies in 339,224 individuals [14], which contains a total of about 2.6 million SNPs. The data set was downloaded from the Genome-scale Integrated Analysis of gene Networks in Tissues Consortium (https://www.broadinstitute.org/collaboration/giant/index.php/GIANT_consortium_data_files#GIANT_consortium_2012-2015_GWAS_Metadata_is_Available_Here_for_Download).

GWAS data sets manipulation

The T2D and BMI GWAS data sets of summary statistics contain a total of about 2.9 and 2.6 million SNPs, respectively. As required by the GPA model, only 2,514,745 overlapped SNPs and the corresponding p values in both traits were used for further analyses. The overlapped SNPs make up a large proportion and contain almost all the potentially significant SNPs (p < 0.05) in each single GWAS data set.

GPA model and pleiotropic SNPs evaluation

In this study, we used the GWAS data sets for two traits: T2D and obesity. Based on the hypothesis testing of genome-wide SNPs, we used their p values (summary statistics for the association testing of individual SNPs) from these two data sets. GPA assumes that the p values from null SNPs (SNPs without true effects) and non-null SNPs follow the uniform distribution and the beta distribution, respectively. Then, we fit the GPA model and evaluated the pleiotropy of SNPs by implementing association mapping with appropriate command. Comparing with similar methods, including conditional false discovery rates (cFDR) and linear mixed models, there are three main advantages (innovations) of GPA method: (1) GPA only needs the summary statistics (p values) as its input which could simplify the analysis procedure and make it popularly applicable. (2) GPA provides more accurate, powerful, and deeper analysis to evaluate local FDR for each SNP, the proportion of SNPs associated with phenotypes, and the overall effect strength of the associated SNPs. (3) With strictly statistical control and integrative consideration of linkage disequilibrium and expression quantitative trait loci, GPA model could accurately and robustly detected risk pleiotropic SNPs for corresponding phenotypes. A detailed description of the GPA model can be found in the original publication [16].

SNPs further prioritized based on false discovery rates (FDR)

Pleiotropic SNPs can be further prioritized based on their local FDR. For a single GWAS, the FDR is defined as the probability that the jth SNP belongs to the null group given its p value:

$$\text{FDR}\left(P_j\right)=\text{Pr}\left(Z_{j0}=1|P_j\right).$$ (1)

For joint analysis of two GWAS data sets, the pleiotropic SNP is defined as the jth SNP associated with both phenotypes:

$${\text{FDR}}_{1,2}\left(P_{j1},P_{j2}\right)=\mathrm{Pr}\left(Z_{j00}+Z_{j10}+Z_{j00}=1|P_{j1},P_{j2}\right).$$ (2)

The specific description of FDR calculation was presented in the original paper [16].

Pleiotropic SNPs annotation

The pleiotropic SNPs were annotated using the online SNP and CNV Annotation Database (SCAN) (http://www.scandb.org/newinterface/index.html). The online annotation tool provided detailed information including chromosome, position, alleles, and genes for each specific SNP.

Gene-based functional annotation, pathway analysis, and gene-disease association analysis

Based on those genes corresponding to pleiotropic SNPs, we performed gene enrichment and functional annotation analysis in the DAVID online system. A total of 227 pleiotropic genes were imported into the online tool for functional annotation and gene enrichment analysis. Comprehensive functional information for all of these genes enriched in specific terms was obtained. In addition, we performed pathway analysis based on all pleiotropic genes using the KEGG database. For the most significant pleiotropic genes (FDR < 0.01), we performed gene interaction analysis and constructed their interaction network using the STRING online tool. Furthermore, gene-disease associations have been observed based on DisGeNET database (https://www.disgenet.org/), the network has been constructed using Cytoscape. Specific descriptions of pathway and network analysis can be found in our previous publication [22, 23].

Expression data analysis of pleiotropic genes

We further explored the potential functions of the identified pleiotropic genes in a comprehensive gene expression data set containing 200 subjects [24]. The data set was downloaded from Gene Expression Omnibus with the accession number GSE32512. Total RNA was isolated from all samples. Transcriptional profiling was performed based on Illumina Human HT-12 v3 Expression BeadChip. The correlations between gene expression levels and different phenotypes including BMI and insulin activities (insulin resistance and insulin sensitivity) were evaluated [24]. We assessed the associations between the pleiotropic genes and the BMI and insulin resistance phenotypes to determine whether the pleiotropic genes identified in our study are potentially associated with both obesity and T2D phenotypes in gene expression data.

Results

Pleiotropic SNPs and further prioritized base on FDR value

Based on the pleiotropy estimation and FDR calculation, 5182 SNPs were found to be related to both T2D and BMI. All of the SNPs can be further prioritized by setting FDR q values < 0.2 as a significance cutoff. In all, 2277 out of 5182 SNPs showed significant pleiotropy in both traits. The distribution of all the pleiotropic SNPs is shown in the Manhattan plot using FDR value as the Y-axis (Fig. 1).

Fig. 1.

Manhattan plot of pleiotropic SNPs. In all, 5182 SNPs showed pleiotropy in T2D and BMI. Two cutoff line: blue line (FDR = 0.2, the recommended cutoff line in GPA model); red line (FDR = 0.01)

Pleiotropic SNPs annotation

In all, 5182 pleiotropic SNPs were annotated to a total of 467 genes. Corresponding to the 2227 significant pleiotropic SNPs (FDR < 0.2), 217 genes were identified as pleiotropic genes (Supplementary Data set 1). For the original GWAS data sets in individual traits, 549 and 2051 SNPs corresponding to 55 and 124 genes reached significance in GWAS level (p < 5E–8) for T2D and BMI, respectively. In our pleiotropic analysis, 2277 SNPs corresponding to 217 genes were identified as pleiotropic SNPs/genes (FDR < 0.2). By comparing with the original results, we found that within the 217 pleiotropic genes, 30 are overlapped with the original GWAS results and 187 are newly identified pleiotropic genes (Fig. 2).

Fig. 2.

Comparison with original GWAS in individual traits. By comparing our results with original GWAS in single trait, 30 genes are significant in our study and original studies and 187 genes are newly identified pleiotropic genes

Integrated gene set enrichment, pathway analysis, and gene-disease association analysis

Corresponding to the total 2277 pleiotropic SNPs, 217 pleiotropic genes were imported into the DAVID online tool for functional annotation and gene enrichment analysis. With an integrated consideration of statistical results and physiological features, we listed the top 10 terms from biological function classification (Table 1). As expected, most of the pleiotropic genes identified in this study were enriched in metabolic-related terms such as “proteolysis,” “positive regulation of macromolecule metabolic process,” and “positive regulation of biosynthetic process.” Additionally, we evaluated the associations between pleiotropic genes and genetic diseases. We list all related diseases with p < 0.05 in Table 2. Six out of 10 terms are associated with diabetes or obesity. Eighteen pleiotropic genes are enriched in T2D (p = 3.82E–06) and 7 genes are enriched in obesity (p = 1.91E–02). Several genes such as SREBF1, CRHR2, FTO, and PPARG show pleiotropy in both traits. Pathway analysis based on the KEGG database indicated that numerous genes including CRHR2 and MAPK1 are involved in metabolism-related pathways such as “type 2 diabetes mellitus,” “ubiquitin mediated proteolysis,” and “maturity onset diabetes of the young” (Table 3). Network analysis and reconstruction based on the most significant pleiotropic genes (FDR < 0.01) provided insight into the interactions between those genes (Fig. 3). Gene-disease association study indicated that 91, 78, and 48 pleiotropic genes are associated with obesity, diabetes, and both syndromes, respectively (Fig. 4). The results supported that pleiotropic analysis using GPA method could reveal more novel and biologically significant genes involved in multiple syndromes.

Table 1.

Top 10 biological function terms (count of enriched pleiotropic gene)

Terms	Count	p value
GO:0006508: proteolysis	18	6.51E–02
GO:0042592: homeostatic process	17	7.56E–03
GO:0006811: ion transport	17	9.27E–03
GO:0010604: positive regulation of macromolecule metabolic process	17	2.38E–02
GO:0048878: chemical homeostasis	15	1.37E–03
GO:0010557: positive regulation of macromolecule biosynthetic process	14	2.59E–02
GO:0031328: positive regulation of cellular biosynthetic process	14	3.57E–02
GO:0009891: positive regulation of biosynthetic process	14	3.94E–02
GO:0006812: cation transport	13	1.74E–02
GO:0045941: positive regulation of transcription	13	2.00E–02

Based on the count of enriched pleiotropic genes, the top 10 biological function terms and their corresponding p values are listed here

Table 2.

Genetic association of diseases/traits and corresponding genes

Term	Count	p value	Genes
Diabetes, type 2	18	3.82E–06	SREBF1, HNF1B, KIF11, WFS1, LEPR, PPARG, IDE, FTO, IGF2BP2, CDKAL1, KCNJ11, TCF7L2, CRHR2, HNF4A, SLC2A2, GIPR, SLC30A8, ABCC8
Obesity	7	1.91E–02	SREBF1, CRHR2, GCKR, LEPR, FTO, PPARG, MC4R
Insulin	6	8.78E–04	LEPR, PPARG, IDE, KCNJ11, ABCC8, TCF7L2
Diabetes, gestational	5	5.79E–04	HNF4A, PPARG, KCNJ11, ABCC8, TCF7L2
Psoriasis	5	2.85E–02	GSTM1, CCHCR1, POU5F1, ZNF148, PPARG
Diabetes, type 2 triglycerides	3	5.97E–03	IGF2BP2, CDKAL1, SLC30A8
Laryngeal cancer	3	1.24E–02	GSTM1, CCND1, BCL2
Binge eating	2	3.47E–02	LEPR, MC4R
Diabetes, type 2 glucose insulin	2	3.47E–02	HNF4A, TCF7L2

The significant diseases (p < 0.05) are listed and sorted by the count of enriched pleiotropic genes. The corresponding p values and the specific genes are shown in the third and fourth columns. The pleiotropic genes are enriched in two major diseases: diabetes and obesity. Some genes such as “SREBF1,” “CRHR2,” “LEPR,” “FTO,” and “PPARG” are involved in both diseases

Table 3.

Pathway analysis based on pleiotropic genes

Pathway	Count	p value	Genes
hsa04080: neuroactive ligand-receptor interaction	8	2.53E–02	CRHR2, PTGER3, OPRL1, LEPR, GIPR, MTNR1B, MC4R, GRIN3A
hsa04930: type 2 diabetes mellitus	5	1.89E–03	MAPK1, SLC2A2, CACNA1G, KCNJ11, ABCC8
hsa04120: ubiquitin mediated proteolysis	5	7.00E–02	UBE2D3, FBXW7, UBE3C, UBE2E2, UBE2R2
hsa05216: thyroid cancer	4	4.17E–03	MAPK1, CCND1, PPARG, TCF7L2
hsa04520: adherens junction	4	5.69E–02	PTPRJ, MAPK1, PVRL2, TCF7L2
hsa05210: colorectal cancer	4	7.02E–02	MAPK1, CCND1, BCL2, TCF7L2
hsa05215: prostate cancer	4	8.04E–02	MAPK1, CCND1, BCL2, TCF7L2
hsa04950: maturity onset diabetes of the young	3	3.24E–02	HNF1B, HNF4A, SLC2A2
hsa05219: bladder cancer	3	8.24E–02	RPS6KA5, MAPK1, CCND1

All pathways are listed and sorted by the count of enriched pleiotropic genes. The corresponding p value and the specific genes are shown in the third and fourth columns. Most are metabolism-related pathways

Fig. 3.

Interactions between significant pleiotropic genes. Based on the STRING database, different interactions between significant pleiotropic genes (FDR < 0.01) are shown in colored lines

Fig. 4.

Gene-disease association study based on pleiotropic genes. In all, 78 (A), 91 (B) and 48 (C) pleiotropic genes are associated with diabetes, obesity and both syndromes base on DisGeNET database

Additional evidence from gene expression data

We investigated the association of 217 pleiotropic genes identified in our study with the phenotypes BMI and insulin resistance in the comprehensive gene expression data. Thirtyseven and 44 pleiotropic genes show significant association with BMI and insulin resistance, respectively (p < 0.05). Among them, 29 genes overlap, indicating that they are significantly associated with both BMI and insulin resistance (Table 4). The additional evidence from expression data analysis of pleiotropic genes strongly supports that the pleiotropic genes including JAZF1 (FDR = 0.0002), SYN2 (FDR = 0.0018), CDKN2C (0.0227), RABGAP1 (0.0295), and UBE2E2 (FDR = 0.0459) and the corresponding SNPs rs849135, rs2119812, rs12855, rs587364, rs17012823 may play crucial roles in both obesity and T2D pathogenesis.

Table 4.

Expression data analysis of pleiotropic genes

	Pleiotropic analysis		BMI		Insulin resistant
Gene	SNP	FDR	logFC	p value	logFC	p value
JAZF1	rs849135	0.0002	1.76	1.39E–05	0.35	2.13E–05
SYN2	rs2119812	0.0018	−1.99	5.49E–03	−0.50	5.27E–04
CDKN2C	rs12855	0.0227	−5.57	1.99E–12	−1.06	6.27E–11
RABGAP1	rs587364	0.0295	−0.80	3.08E–03	−0.15	7.46E–03
UBE2E2	rs17012823	0.0459	0.99	1.83E–04	0.14	8.22E–03
ATP2B4	rs17537593	0.0479	−1.36	4.51E–03	−0.38	7.26E–05
DDX42	rs11079511	0.0554	−0.48	2.87E–02	−0.14	1.67E–03
EIF3H	rs12677487	0.0600	−0.96	4.42E–05	−0.22	2.13E–06
NUCB2	rs214105	0.0699	1.81	1.91E–05	0.39	5.18E–06
SKIV2L	rs437179	0.0759	−0.77	3.38E–03	−0.13	1.20E–02
BPTF	rs12449442	0.0839	0.94	3.65E–03	0.14	3.90E–02
SENP2	rs3087964	0.0876	−1.92	2.05E–06	−0.40	8.84E–07
WDR68	rs17631394	0.0899	−0.74	3.16E–02	−0.19	7.49E–03
SOBP	rs3923954	0.1048	2.15	8.65E–06	0.53	3.62E–08
SLIT3	rs10069222	0.1096	0.78	4.54E–02	0.25	1.85E–03
THADA	rs17334919	0.1108	1.66	2.17E–05	0.41	1.33E–07
SLC39A11	rs9912666	0.1145	2.08	3.78E–06	0.43	2.84E–06
CCND1	rs649392	0.1228	4.27	1.86E–09	0.78	8.58E–08
BCL2	rs12454712	0.1326	−1.04	2.37E–03	−0.26	1.72E–04
OGDH	rs6961567	0.1342	−1.22	1.38E–05	−0.12	3.27E–02
GALNT10	rs6860445	0.1425	1.68	4.72E–05	0.29	4.68E–04
PROS1	rs8178610	0.1463	2.24	3.03E–10	0.45	3.39E–10
PHACTR2	rs1082	0.1543	0.64	3.10E–02	0.14	1.55E–02
NOPE	rs2289045	0.1570	−2.57	1.68E–05	−0.43	4.02E–04
HLA-DRA	rs3129882	0.1588	3.33	5.45E–09	0.63	7.17E–08
FBXW7	rs1516822	0.1633	1.42	3.96E–05	0.21	3.24E–03
C15orf44	rs12909511	0.1662	−1.50	2.58E–03	−0.28	4.84E–03
PLEKHM2	rs16851970	0.1771	0.82	5.05E–04	0.17	3.76E–04
TOM1L2	rs7222480	0.1872	−2.73	3.83E–08	−0.52	3.18E–07

The first column lists the 29 genes that show evidence of pleiotropy in both our study (FDR < 0.2) and the validation gene expression data (p < 0.05). The second and third columns list the corresponding SNPs and FDR value in our GWA-based pleiotropic analysis. The last four columns list the log fold change (FC) and p values in gene expression data which may reveal the association between genes and the phenotypes BMI and insulin resistance, respectively

Discussion

It is well known that T2D and obesity are physiologically related diseases that may influence each other in multiple ways. Both conditions are global public health problems. Despite many studies focused on the pathogenesis of T2D and obesity [25–27], either in a single gene, pathway or in various omics investigations, genetic intercommunity studies between these two diseases are very limited. With the assumption that two closely related traits are likely to share common genetic factors [28, 29], we performed pleiotropic analysis based on large GWAS data sets regarding T2D and BMI. Although several methods [16, 30, 31] for pleiotropy analysis between two GWAS summary results, GPA has better performance in statistical power for identifying disease-associated genes than many other approaches including cFDR [16].

Since the purpose of GWA-based pleiotropic analysis is to identify pleiotropic genes which play significant roles in both traits, we compared the results between our pleiotropic analysis and the original GWAS in single phenotype. As described in the “Results,” 2227 pleiotropic SNPs corresponding to 217 unique pleiotropic genes were identified in our study. Among these genes, 30 are overlapped and significant in GWA level (p < 5E–8), and 187 are newfound genes which may be ignored in the original GWAS studies. We do not intend to compare the number of significant SNPs or genes between multiple studies based on different statistical methods and values (e.g., p value and FDR value). However, pleiotropic analysis in our study indeed sheds light on those causal SNPs and genes which may be ignored in single GWAS. For instance, we verified multiple SNPs including rs4506565 (FDR = 1.23E–08), rs12255372 (1.74E–07), rs1558902 (7.23E–10), rs1421085 (FDR = 7.58E–07) and the corresponding genes TCF7L2 and FTO which are significant in GWA level in both original studies. More importantly, beyond verification, we highlighted several SNPs such as rs864745 (FDR = 3.96E–04), rs17175643 (FDR = 1.05E–04), rs11603334 (FDR = 1.9E–04) and the corresponding genes JAZF1, MC4R, CENTD2 which do not reach genome-wide significance in GWA level in either or both of the original data sets, but show significant pleiotropy in our study.

GWAS is a popular approach for systematical biological investigation. With high throughput and coverage, it is able to identify millions of SNPs in a large amount of subjects [32, 33]. SNPs could affect specific phenotypes in multiple ways, including regulating gene expression level [34, 35]. In order to gain additional evidence, we performed expression data analysis of pleiotropic genes. By evaluating the correlation between genes and the phenotypes BMI and insulin resistance (an important phenotype of T2D), several pleiotropic genes such JAZF1, SYN2, and CDKN2C showed significant association with both phenotypes. Several previous epidemiologic and genomic studies demonstrated a close relationship between JAZF1 and T2D [36–38]. Meanwhile, numerous functional studies in multiple tissues have shown that JAZF1 can reduce lipid synthesis and increase lipolysis in adipocytes and liver cells, which may represent a potential target in intervention and treatment of T2D and obesity [39–41]. A number of previous publications have suggested crucial roles for genes SYN2 and CDKN2C in T2D and obesity [42–45]. As we know, the variance of gene expression in different individuals may lead to different expression of corresponding proteins, thus influencing specific phenotypes. Evidence from gene expression data provided strong and direct support for our pleiotropic analysis using the GPA method.

FTO is a nuclear protein coding gene, with the full name “Alpha-Ketoglutarate Dependent Dioxygenase.” RNA (mRNA, tRNA, and snRNA) demethylase mediated by FTO play a crucial role in the physiology of energy and material metabolism including adipogenesis and glycometabolism [46, 47]. Previous studies suggested that FTO could regulate the fat mass and body weight by mediating the thermogenesis and adipocyte differentiation into brown and white fat cells [48]. TCF7L2 play a significant role in Wnt signaling pathway. It encodes a high-mobility group box-containing transcription factor which is involved in blood glucose homeostasis and adipogenesis. As a well-known T2D candidate gene, TCF7L2 is also a crucial transcriptional effector in Wnt signaling pathway which could significantly regulate the development and function of adipocytes [49]. Our integrative pathway and network analysis suggested that both TFO and TCF7L2 were involved in metabolism-related pathways or diseases such as “glucose homeostasis,” “carbohydrate homeostasis,” “diabetes mellitus,” and “type 2 diabetes.”

We note that the imbalance of data structure inherent in performing a study between two GWAS data sets may lead to unexpected false positive results. Generally, two GWAS data sets with uncoordinated sample size and/or count of SNPs are not the ideal data sets for pleiotropic studies by the GPA method [16]. To avoid missing important information, the ideal scenario is to use all SNPs and the corresponding statistical values in both data sets for pleiotropic evaluation. However, overlapped SNPs are required by most methods including GPA. Therefore, among all the conditions, a comparative count of total SNPs is the most important element between the two data sets. In our study, although the sample sizes are different, the total counts of SNPs between T2D and BMI data sets are comparable. Furthermore, by comparing with the original data sets, the overlapped SNPs we used for pleiotropic analysis make up a large proportion of total SNPs (around 90%) and cover almost all significant SNPs (p < 0.05) in both original GWAS data sets. Based on data quality control, we can effectively minimize the false positive rate and enhance the reliability of our results. Therefore, imbalance of sample size in the two GWAS data sets that we used may have minimum impact for unexpected false positive results in our study.