RESISTANCE DEVELOPMENT

The present invention pertains to different genetic markers of importance to the molecular mechanism involved in insulin resistance. A number of SNPs (single nucleotide polymorphisms) that are associated with insulin resistance have been located in the gene vesicle associated membrane protein-associated protein A (VAPA). Individual responses to a dietary challenge are expected to vary among individuals. Individuals with either a weak or strong response in insulin resistance upon dietary changes in glycemic load showed distinct genotype profiles. These markers have been extensively screened and connections between variation in SNPs and changes in insulin resistance in response to diets with different glycemic load have been identified.

An association between genetic variability in VAPA and insulin resistance has been found where several specific SNPs on identified quantitative trait loci (QTLs) are pinpointed.

Susceptibility loci traits for insulin resistance and SNPs which are involved in the molecular mechanism of the VAPA genetic interactions with insulin resistance have been identified. The protein encoded by this gene is a type IV membrane protein. It is present in the plasma membrane and intracellular vesicles. It may also be associated with the cytoskeleton. This protein may function in vesicle trafficking, membrane fusion, protein complex assembly and cell motility. Alternative splicing occurs at this locus and two transcript variants encoding distinct isoforms have been identified.

One aspect of the present invention is directed to specific SNPs as new markers of candidate QTLs related to genetic aspects of developing insulin resistance. Another aspect of the present invention involves the use of VAPA and plasma protein inhibitor of activated STAT-1 (PIAS 1) as candidate genes for molecular mechanisms involved in insulin resistance. Yet another aspect of the present invention involves a specific marker SNP in the GIP (gastric inhibitory polypeptide) gene, a candidate expressional QTL (eQTL) affecting plasma plasminogen activator inhibitor-1 (PAI-1) concentrations related to insulin resistance.

The identified genetic markers can be used in the diagnosis of insulin resistance correlated with dietary diseases, especially glycemic loads. Furthermore such markers can be used in developing suitable drugs for regulating glycemic response in people with such diseases. Furthermore, such markers associated with insulin resistance can be used to explain individual physiological responses to dietary glycemic load. SNP typing can be used to provide concrete dietary advice to persons genetically predisposed to type II diabetes (T2D).

BACKGROUND

Type 2 diabetes (T2D) is defined as chronic hyperglycemia, manifested when insulin production is overwhelmed by insulin resistance in target cells, leading to a decreased ability of glucose uptake (Tripathy and Chavez, Curr Diab Rep, 2010, 10(3): pp. 184-91, incorporated herein by reference). Insulin resistance, however, precedes the onset of T2D by many years (Pagel-Langenickel et al., Endocr Rev, 2010, 31 (1): pp. 25-51, incorporated herein by reference), and in addition to be a risk factor for T2D it is also an independent predictor for e.g. hypertension, coronary heart disease (CHD), stroke, and cancer (Facchini et al., J Clin Endocrinol Metab, 2001, 86(8): pp. 3574-8, incorporated herein by reference). Even though obesity is associated with increased insulin resistance, individuals of normal weight do also experience variable sensitivity to insulin (McLaughlin et al., Metabolism, 2004, 53(4): pp. 495-9, incorporated herein by reference).

Already 30 years ago it was stated that the prevalence of T2D could be reduced by lifestyle changes, but so far the incidence of T2D has only been increasing, and the expansion is now called a modern epidemic (Meigs, Diabetes Care, 2010, 33(8): pp. 1865-71, incorporated herein by reference). There are at least two plausible explanations for this: Firstly, the dietary guidelines may be underestimating the influence of dietary glycemic load on hyperinsulinemia (Ludwig, Jama, 2002, 287(18): pp. 2414-23, incorporated herein by reference). Secondly, the same guidelines may be too general. The capability to study the complex genetics behind interindividual metabolic differences (Lairon et al., Public Health Nutr, 2009, 12(9 A): pp. 1601-6, incorporated herein by reference) has been developed only recently, revealing benefits of personalized nutrition among high-risk persons (Kaput, J., Curr Opin Biotechnol, 2008, 19(2): pp. 1 10-20; Martinez et al., Asia Pac J Clin Nutr, 2008, 17 Suppl 1 : p. 1 19-22; both incorporated herein by reference).

Insulin resistance is a pathophysiological trait characterised by an aberrant blood lipid profile, endothelial dysfunction, increased plasma concentration of procoagulant factors, and markers of inflammation (Goldberg, R.B., J Clin Endocrinol Metab, 2009, 94(9): pp. 3171- 82, incorporated herein by reference). The etiology of insulin resistance is complex and unlikely to be the same in every individual. A major determinant, though, seems to be cytokine induced activation of proinflammatory pathways in insulin target cells, reducing insulin sensitivity. This activates and attracts immune cells, and establishes a feed forward loop resulting in macrophage infiltration of the tissue, and additional cytokine secretion (Olefsky and Glass, Annu Rev Physiol, 2010, 72: pp. 219-46, incorporated herein by reference). The inflammatory origin can be retraced to cellular stress, caused by metabolic imbalance, hence, called metaflammation (Hotamisligil, Nature, 2006, 444(7121): pp. 860-7, incorporated herein by reference). Prolonged malnutrition leads to chronic metaflammation, and may eventually cause degeneration of tissue, and onset of disease (Kushner et al., Arthritis Care Res (Hoboken), 2010, 62(4): pp. 442-6, incorporated herein by reference). Hyperglycaemia and hyperinsulinemia following a meal rich in easily digested carbohydrates are associated with cellular stress and increase of inflammatory markers (O'Keefe et al., J Am Coll Cardiol, 2008, 51(3): pp. 249-55, incorporated herein by reference). Diets with low glycemic load and glycemic index are suggested to silence metaflammation, and subsequently increase insulin sensitivity (Barclay et al., Am J Clin Nutr, 2008, 87(3): pp. 627-37; McKeown et al., Diabetes Care, 2004, 27(2): pp. 538-46; and Qi and Hu, Curr Opin Lipidol, 2007, 18(1): pp. 3-8; all incorporated herein by reference).

Current evidence suggests that insulin resistance and the associated abnormalities constitute complex phenotypes, explained by both environmental and genetic factors. The genetic makeup underlying these traits consists of several quantitative trait loci (QTL), whereof each QTL only explains a small fraction of the phenotype. The limited effects of these individual QTL make them difficult to identify, but the list of allelic variants associated with susceptibility to T2D development, in terms of single nucleotide polymorphisms (SNPs), is growing (Voight, B.F., et al., Twelve type 2 diabetes susceptibility loci identified through large-scale association analysis. Nat Genet, 2010. 42(7): p. 579-89, incorporated herein by reference). Also SNPs associated directly with insulin resistance have been found, but this line of research is in an early phase. (See Kantartzis et al., Clin Sci (Lond), 2009, 1 16(6): pp. 531-7; Liu et al., J Clin Endocrinol Metab, 2009, 94(9): pp. 3575-82; Palmer et al., Diabetes, 2004, 53(11): pp. 3013-9; Richardson et al., Diabetologia, 2006, 49(10): pp. 2317-28; Ruchat et al., Diabet Med, 2008, 25(4): pp. 400-6; and Smith et al., Diabetes, 2003, 52(7): pp. 161 1-8; all incorporated herein by reference.) The expression of a gene is the most basic phenotype in an organism. The genotype determines complex phenotypic traits through expression of several genes: expressional QTL (eQTL) (Jansen and Nap, Trends Genet, 2001, 17(7): pp. 388-91 ; and Schadt et al., Nature, 2003, 422(6929): pp. 297-302, both incorporated herein by reference). eQTL provide a direct link between genotype variation and gene- or pathway activities. The motivation to study how SNPs associated with a disease or a phenotypic trait may affect gene expression is to gain a direct understanding of the molecular mechanisms affected by the allelic variation (Rockman and Kruglyak, Nat Rev Genet, 2006, 7(1 1): pp. 862-72, incorporated herein by reference).

Homeostatic model assessment (HOMA) is a method for assessing surrogate measures of pancreatic β-cell function, insulin sensitivity, and insulin resistance derived from fasting blood glucose and insulin, alternatively insulin connecting peptide (C-peptide) concentrations (Wallace et al., Diabetes Care, 2004, 27(6): pp. 1487-95, incorporated herein by reference). The model was first proposed in 1985 (Matthews, et al., Diabetologia, 1985, 28(7): pp. 412-9, incorporated herein by reference), and an updated computer model (HOMA2) was published in 1998 (Levy et al., Diabetes Care, 1998, 21(12): pp. 2191 -2, incorporated herein by reference). The calculation of insulin resistance designated as HOMA2 IR, is calibrated to a reference population, where the value 1 is set as normal (Wallace et al., 2004). HOMA2 IR was found to be a significant determinant of insulin resistance (Mojiminiyi et al., Clin Chem Lab Med, 2010, incorporated herein by reference).

In the present study, performed on modestly overweight but otherwise healthy individuals, associations between variation in SNPs and changes in insulin resistance in response to diets with different glycemic load were examined. SNPs were linked to genes and biological functions to develop an understanding of the molecular mechanisms potentially involved in onset of insulin resistance.

METHODS

Subjects and study outline

A randomized, controlled cross-over diet intervention trial was conducted on thirty- two young and healthy women and men, with body mass index (BMI, in kg/m ² ) between 24.5 and 27.5. Iso- and normocaloric meal replacement diets (MRDs) constituted all nutrients consumed during the study periods of two times six days with an eight day wash-out period in-between. Fasting blood samples were collected before and after each diet period, and effects of dietary intake on leukocyte gene expression profiles and insulin resistance were analyzed, as described previously ((Arbo I, Brattbakk HR, Langaas M et al. A balanced macronutrient diet induces changes in a host of pro-inflammatory biomarkers, rendering a more healthy phenotype; a randomized cross-over trial, 2010, (manuscript submitted), incorporated herein by reference). The two MRDs were: a high-carbohydrate diet (AHC) composed of 65: 15:20 energy percent (E %) of carbohydrates, proteins, and fats; and a moderate-carbohydrate diet (BMC) with 27:30:43 E % of carbohydrates, proteins, and fats. The glycemic load of the AHC diet was calculated to be 2.71 times higher than the BMC diet.

Data extracted from samples were grouped and coded, according to diet and time of sampling. The abbreviations AHC0, AHC6, BMC0, and BMC6 denote before (day 0) and after (day 6) the AHC and the BMC diet intervention, respectively. Pair-wise analyses of data were performed for four different comparisons, which will be referred to throughout this paper: 1 ) AHC6-AHC0 and 2) BMC6-BMC0 identified responses to the AHC and the BMC diets, respectively, during six days on the respective diets. The comparison 3) BMC6-AHC6 identified the differences between the end-point responses to diet AHC and BMC after six days on diet, and finally, 4) (BMC6-BMC0)-(AHC6-AHC0) identified differences between the responses to AHC and BMC dieting. Complementary and more detailed information about subject recruitment, exclusion criteria, subject baseline characteristics, MRD compositions, and sampling techniques were described previously (Arbo I, Brattbakk HR, Langaas M et al. A balanced macronutrient diet induces changes in a host of proinflammatory biomarkers, rendering a more healthy phenotype; a randomized cross-over trial, 2010, {manuscript submitted), incorporated herein by reference).

Microarrav hybridization and data analysis

Microarray analysis and preprocessing of microarray data was performed as previously described (Arbo I, Brattbakk HR, Langaas M et al. A balanced macronutrient diet induces changes in a host of pro-inflammatory biomarkers, rendering a more healthy phenotype; a randomized cross-over trial, 2010, (manuscript submitted), incorporated herein by reference). Briefly, leukocyte gene expression profiling was done on the HumanHT-12 Expression BeadChip v3.0 (Illumina). After removal of two outlier samples, background correction based on negative controls, quantile-quantile normalization, signal log ₂ - transformation, and removal of not detected or bad probes, 27 372 unique probes were left in the "gene expression dataset". The paired analyses of AHC6-AHC0 and BMC6-BMC0 identified 3225 and 1370 differentially expressed genes, respectively, where 843 genes overlapped between the analyses. For the paired groups BMC6-AHC6 and (BMC6-BMC0)- (AHC6-AHC0), no differentially expressed genes were identified. Microarray data were submitted to ArrayExpress (www.ebi.ac.uk/arrayexpress, accession number: E-TABM-1073).

Analysis of the Bio-Plex Diabetes Panel and assessment of insulin resistance

Protein concentration analyses and assessment of insulin resistance were performed as previously described (Arbo I, Brattbakk HR, Langaas M et al. A balanced macronutrient diet induces changes in a host of pro-inflammatory biomarkers, rendering a more healthy phenotype; a randomized cross-over trial, 2010, (manuscript submitted), incorporated herein by reference), using fasting EDTA-plasma samples. Bio-Plex Diabetes Panel assays (Bio-Rad Laboratories Inc., Hercules, CA, USA) were performed using Luminex xMAP™ technology, with a Bio-Plex 200 suspension array reader, and the data was extracted with the Bio-Plex Manager 5.0 software (Bio-Rad Laboratories Inc.). Briefly, analysis of the paired groups showed decreased (P < 0.05) plasma concentrations of visfatin (nicotinamide phosphoribosyltransferase, Nampt), and increased plasma concentration of resistin (P < 0.05) during AHC dieting. Likewise, during BMC dieting the analysis showed decreased plasma concentrations of insulin, C-peptide, glucagon, plasminogen activator inhibitor- 1 (PAI-1), glucagon-like peptide-1 (GLP-1), tumor necrosis factor-a (TNF), interleukin-6 (IL-6), and visfatin, and increased plasma concentrations of resistin. Gastric inhibitory polypeptide (GIP), ghrelin, and leptin did not respond to any of the diet interventions.

The HOMA2 calculator version 2.2.2 ^® (Diabetes Trials Units, University of Oxford, www.dtu.ox.ac.uk/homacalculator/index.php) (Matthews et al., 1985) was used to determine changes in insulin resistance in terms of HOMA2 IR. There was an average decrease in HOMA2 IR during both the AHC diet and the BMC diet, but the downregulation was only significant during BMC dieting.

Genotyping

DNA was extracted from EDTA-blood using E.Z.N.A Blood DNA Kit (D3392, OMEGA Bio-Tek, Inc., Norcross, GA, USA). The subjects were genotyped using the -200 K Cardio-MetaboChip (Metabochip) SNP array , an Infinium iSelect HD Custom Genotyping BeadChip (Illumina, San Diego, CA, USA), designed by the Cardio-MetaboChip Consortium (Broad Institute, Cambridge, MA, USA), and analyzed according to the Infinium HD Assay Ultra, Manual Experienced User Card. The Metabochip consists of SNPs associated with diseases or traits relevant to metabolic and atherosclerosis-cardiovascular endpoints, including T2D and hyperglycemia. The BeadChips were read by a BeadArray™ reader, and data were exported to GenomeStudio™ V2009, Genotyping VI .1.9 (Illumina), for visual quality control of genotype clustering, and extraction of quality measures (ChiTestl OO and GenTrain Score) (Illumina, GenomeStudioTM Genotyping Module vl. O User Guide. 2008, Illumina, Inc: San Diego, CA, incorporated herein by reference). The ChiTestlOO is a p-value calculated for each SNP, reflecting the deviation of that SNP to the genotype distribution according to the Hardy- Weinberg Equilibrium (HWE), using the χ ² statistic, normalized to 100 subjects. GenTrain Score is a measure of SNP clustering performance indicated by a number increasing with cluster quality, form 0 to 1.

Candidate gene selection

A set of 22 transcription regulators and seven ligand-dependent nuclear receptors central to insulin resistance development (Olefsky and Glass, 2010; Hotamisligil, 2006; and Wymann and Schneiter, Nat Rev Mol Cell Biol, 2008, 9(2): pp. 162-76; all incorporated herein by reference) were selected. The selected candidate genes were uploaded to the Ingenuity Pathway Analysis 8.7 (IPA Ingenuity Systems®, Redwood City, CA, USA, www.ingenuity.com) to find the upstream activators and inhibitors, and downstream target genes of the transcription regulators and the nuclear receptors. No filters were applied in IPA regarding species, tissues or cell lines, but an upper limit of 150 upstream and 150 downstream genes was defined. The SNPs linked to the extended selected list of 276 candidate genes were extracted from the dbSNP database (www.ncbi.nlm.nih.gov/projects/SNP, National Center for Biotechnology Information, U.S. National Library of Medicine, Bethesda), and matched with 469 SNPs on the Metabochip. These 469 SNPs (linked with 276 candidate genes) were uploaded to the web server FASTSNP (Yuan et al., Nucleic Acids Res, 2006, 34 (Web Server issue): pp. W635-41, incorporated herein by reference) to prioritize the SNPs that were most likely to have functional effect on the expression of the linked gene. According to a decision tree, each SNP was assigned a risk score between 0 and 5. Risk score 0 means that the SNP has no known effect (e.g. located in a downstream or upstream untranslated region, nearby the gene), and 5 means that the SNP has a functional effect (e.g. introduces a stop codon and hence premature translational termination). Basically all SNPs with risk score lower than 2 were discarded. Since several SNPs with risk score 2 or higher were linked to a single gene, we defined an upper limit of seven SNPs per gene. That was done by increasing the risk score claim one factor at the time, until the number of SNPs was at most seven. The result was a list of 190 SNPs.

SNP selection

Four different selections of SNPs were used in the analyses:

1. The ref-SNP selection— 71 061 Metabochip SNPs assigned with a reference SNP ID (rs) with more than one SNP type among the 32 subjects. The ref-SNP selection was used to screen for SNPs that could be associated with HOMA2 IR.

2. The gene-SNP selection— a subset of 23 382 SNPs linked according to the dbSNP database with one or more genes present in the "gene expression dataset". This resulted in 35 082 SNP and gene expression value (log ₂ -ratio) pairs, since several genes were represented with multiple probes on the HumanHT-12 Expression BeadChip. The gene-SNP selection was used to screen for pairs where the SNP was associated with the expression of the gene.

3. The candidate gene-SNP selection— the subset of 190 SNPs that according to the dbSNP database were linked with the genes in the candidate gene list (described above). This resulted in 364 SNP and gene expression value pairs. The candidate gene-SNP selection was used to screen for association between SNPs and HOMA2 IR, and associations between SNPs and gene expression.

4. The diabetes panel-SNP selection— a subset of 7 SNPs that according the dbSNP database were linked with genes coding for the proteins on the diabetes panel. This set of SNP selection was examined for association with the expression of proteins or genes of the diabetes panel. The SNPs were also tested for association with HOMA2 IR. Statistical analyses

For all analyses a two-stage strategy was performed. In the first stage, analysis of variance (ANOVA) was performed to test the null hypothesis, whether there was no difference in either HOMA2 IR, gene expression (log ₂ -ratio), or protein concentration (log _e - ratio) change between the genotypes. Genotype was used as covariate, and changes as response variables. P-values were adjusted for multiple testing using the Benjamini-Hochberg step up algorithm (Benjamini and Hochberg, Journal of the Royal Statistical Society. Series B (Methodological), 1995, 57(1): pp. 289-300, incorporated herein by reference) to control the false discovery rate (FDR). In the second stage, a one-sample, two-sided t-test was assigned to test if the change in HOMA2 IR, gene expression, or protein concentration, was different from zero for any of the genotypes. For the ref-SNP selection and the gene-SNP selection, the second stage was performed only for the 100 best ranked entries, according to the ANOVA p- values. Hence, eight Top 100 lists were generated, one for each comparison, the ref-SNP selection and the gene-SNP selection separately (see Supplementary tables 1 -8). Within these lists the t-test p-values were adjusted for multiple testing using the Benjamini-Hochberg step up algorithm.

Unsupervised hierarchical clustering analyses were performed, using Manhattan distance measures and complete linkage. PCA was performed with discrete data, where the three possible genotypes were represented by numerical values (0, 1, 2). Analyses were performed using the R statistical analysis framework (R Development Core Team, R: A Language and Environment for Statistical Computing, 2010; Available from: www.r- project.org, incorporated herein by reference).

Functional analysis to identify biological functions and diseases significantly associated with gene lists were performed using IPA 8.7 (Ingenuity). Since the Metabochip is custom made, biased by SNPs associated with metabolic and cardiovascular traits, a custom reference set was also used in all analyses. This was composed of all the 10 515 genes that according to the dbSNP database were linked to the 71 061 SNPs on the Metabochip. P- values were adjusted for multiple testing using the Benjamini-Hochberg step up algorithm. RESULTS

SNPs associated with HOMA2 IR

Insulin resistance is a complex trait and the contribution of each single locus to the phenotype is small. To propose loci involved in the manifestation of this trait, the environmental homeostasis was challenged by introducing the subjects to two different diets. The responding change in HOMA2 IR for the four comparisons was related to SNPs in the ref-SNP selection. The biological relevance to insulin resistance was examined for all SNPs with FDR < 0.2, a cut-off used in larger cohorts (> 3000 subjects) earlier (Povel et al., Int J Obes (Lond), 2010, 34(5): pp. 840-5, incorporated herein by reference).

The change in HOMA2 IR during the AHC diet was associated (FDR < 0.1) with four SNPs, with identical allele distribution between the subjects (Figure 1A). The first SNP, rs 16961756 (cgggccttcctcgccagcacctccattcct[a/g]aggctcacgtgggagagacagtgt ggagag)(SEQ ID NO: 1) (Chrl 7: 17359619, G→A) was located 126 base pairs (bp) upstream of a putative pseudogene (LOC I 00288179). This finding is supported by similar allele distribution in the closest neighboring SNP on the Metabochip, rs 1242483 GAAGTTAAGAGTAAATAAAATAGTCA[C/T]GTTTGGGAGCATGAAGGAGCCGGC A (Chrl 7: 17351675, T→C, P = 0.002)(SEQ ID NO: 2)

The three other SNPs,

rs29095 (tccctcgaataaaggtgaaatttttaaaat[a/c]tcagtgaataggaatgtgcaaagc tctaag)(SEQ ID NO: 3) (Chrl 8:9957549),

rs7237794 (ctgcccctcccgacacagcacatacacaca[c/t]tgacgttttgctactacagcatat agcctt)(SEQ ID NO:

4) (Chrl 8:9951304), and

rs917688 (ttctctcatgcttaatatttggaactataa[a/c]gctaaaggccattgacgtagctaa aaatct)(SEQ ID NO:

5) (Chrl 8:9962736), (Chrl 8:9951304..9962736, C→A, T→C, and C→A, respectively) were closely linked, and rs7237794 and rs917688 were located in an intron region, and in the untranslated region of the 3 'end of the gene vesicle-associated membrane protein-associated protein A, 33kDa (VAPA), respectively. The 29 subjects homozygous for the consensus allele had an average downregulation of HOMA2 IR during the AHC diet (estimate of average change (x ) = -0.279, FDR = 0.004), while the three remaining heterozygotes had an average upregulation (3c = 1.000, FDR = 0.098). This response to the AHC diet was only modestly reflected on the VAPA gene expression level. There was no change in HOMA2 IR among the homozygotes ( x = 0.008, P = 0.859), while among the heterozygotes there was a decrease ( 3c = -0.216, P = 0.014).

Another association (FDR < 0.02) was found between HOMA2 IR change during the AHC diet and the SNP rsl 0803976 (Figure IB) (Chr2: 185428946, C→T)(CATTAA

AAGCTATCATCTAACATTGC[C/T]TGGAGTGTTTATTTTTAAGTGCATA)(SEQ ID NO: 6), located 34 Kbp upstream of the nearest gene (zinc finger protein 804A). The 27 individuals homozygous for the consensus allele experienced an average decrease in HOMA2 IR during the AHC diet ( 3c = -0.31 1, FDR = 0.004). The four heterozygotes experienced an average increase during the AHC diet ( 3c = 0.800, FDR = 0.103), and the response difference between the AHC ( 3 = 0.800) diet and the BMC diet (3c = -0.275) was significant ( 3c = - 1.075, FDR = 0.048). Only one was homozygous for the alternative allele.

The same procedure was followed for the candidate gene-SNP list. HOMA2 IR change during the BMC diet was associated with the SNP rs649471 1 (Figure 1C) (aaaattgaggaaaatcccagaagatagagc[c/t]aaaagacaagagatgtaaaaatgc acaagg)(SEQ ID NO: 7) (FDR = 0.047, Chrl 5:68374027, T→C). Those homozygous for this SNP had an average decrease in HOMA2 IR (TT, n = 9, 3c = -0.644, P = 0.004; CC, n = 9, 3c = -0.422, P = 0.003), while the heterozygotes had no significant change (CT, n = 14, 3c = 0.021 , P = 0.773). The SNP rs649471 1 was located in an intron region of the transcription factor protein inhibitor of activated STAT-1 (PIAS1). The nearest neighbouring SNP on the Metabochip, rsl489595 AATTTTCTGTTTACACAAGTGATTCT[A/G]TAAGCAAACCAGGGTTCTCCATGGT (SEQ ID NO: 8) (Chrl5:68377126, A→G, P = 0.014), also located in an intron region of PIAS 1, showed the same changes in HOMA2 IR among hetero- and homozygotes. The genotype specific changes were not reflected in the mRNA expression data of PIAS1.

SNP in the GIP gene is associated with PAI- 1 protein concentration in plasma

To screen for cis- and trans- regulating eQTLs that affected expression of genes central in pathogenesis of T2D, we tested the association between the SNPs in the diabetes panel-SNP and HOMA2 IR. Association was detected for the SNP rs2291726 (Figure ID) (Chrl 7:47039254, C→T)(TCTAGGGACACTTGAATCTTTTAATA[C/T]C

TGAACCCCAAAAGCAGAGGGTACC)(SEQ ID NO: 9) and the protein concentration of PAI-1 in plasma. The SNP was located in an intron region of the gene coding for GIP. For the eight individuals homozygous for the consensus allele, the average protein concentration (log _e -ratio) changed during the AHC diet differed significantly from the BMC diet ( x = - 1.003, P = 0.016). For the 21 heterozygotes and the three homozygous for the alternative allele there were only minor or no differences in PAl-1 concentration changes ( = -0.037, P = 0.695; x = -0.075, P = 0.848, respectively). This suggests that nucleotide variation in GIP mRNA may have downstream effects on protein concentration of PAI-1. However, precaution has to be made interpreting this finding, since the deviation from HWE is significant (P = 0.029).

SNPs associated with HOMA2 IR change are related to type 2 diabetes

Since insulin resistance is manifested by numerous QTL, we wanted to explore how the genotype profile in each individual correlated with the change in HOMA2 IR in response to the diets. Heatmaps were generated showing hierarchical clustering of the ref-SNP selection Topi 00 SNPs associated with HOMA2 IR. SNPs were clustered according to allele distribution, and the subjects were sorted according to HOMA2 IR differences (increasing from left to right, Figure 2A) in the comparison corresponding to the Top 100 list. The genotype profiles for the subjects with the largest increase in HOMA2 IR during the AHC diet were notable (Figure 2A, right). Some clusters of SNPs seem to have a dominant role influencing HOMA2 IR. The genotype profiles of the subjects with the largest decrease in HOMA2 IR during the BMC diet also differed distinctively from the rest of the subjects (Figure 2B, left).

Since the subjects with the strongest increase in HOMA2 IR during the AHC diet had such a distinct genotype profile, these were suspected to be involved in our most significant associations between SNP and HOMA2 IR. To determine if this was due to technical artifacts we generated a dendrogram and a PCA plot (showing the three first principal components) based on allele information from the 71 061 SNPs to examine whether we had outlier individuals (Figure 3). The figures suggest that there were two outliers, subjects 22 and 25, but neither of these contributed to the most significant associations between genotype and HOMA2 IR. However, those who did (especially subject 2, 12, 15, and 28) could not be considered as being outliers in these analysis, clustering well with the other subjects.

To explore functional and biological information about the SNPs that showed the highest association with HOMA2 IR, we assessed IPA's Functional Analysis. We extracted the genes that according the dbSNP database were linked to all the SNPs in the ref-SNP selection Top 100 lists. We found 366 unique SNPs in the four lists, and these were linked with 150 unique genes. The gene set was significantly associated with T2D, displaying an FDR-value equal to 6.39xl0 ^"13 for the sum of all four ToplOO lists (Table 1). The SNPs included in the ref-SNP selection Topl OO lists were also associated with several traits that usually co-exist with insulin resistance, like cardiovascular disorder, hypertension, and immunological disorder.

Genotype specific gene expression changes

To identify potential insulin resistance eQTLs, we matched the genes of the ToplOO pairs from the gene-SNP lists for the four comparisons, with the genes related to insulin resistance in the literature, using the following search in PubMed (NCBI, NIH, USA): ("Diabetes Mellitus, Type 2"[MeSH] OR "Insulin resistance" [MeSH] OR "Hyperglycemia"[MeSH] OR "Insulin-Secreting Cells"[MeSH]). None of the pairs of SNPs and genes related to insulin resistance showed significant association between genotype and expression changes, but several genes showed significant genotype specific expression changes (FDR < 0.05) in response to diet AHC and BMC (Table 2). This suggests that genotype is a considerable variable, contributing to interindividual gene expression variability.

DISCUSSION

In this study we have defined a method to relate SNPs to phenotypic changes in response to an intervention, and applied this method to identify potential susceptibility loci for insulin resistance. The method should also be applicable on larger cohorts. We observed distinctive genotype profiles among strong responders to high and low glycemic load, concerning increase and decrease of insulin resistance, respectively. Several eQTL were found linked to genes related to insulin resistance, showing inter-genotype variability. On a limited number of subjects, we successfully applied statistical and bioinformatical methods new to this area of genetic research.

Our most significant finding is association of insulin resistance to VAPA, a protein previously shown to play a role in the vesicle budding and fusion events involving protein transport in cells (Weir et al., Biochem Biophys Res Commun, 2001 , 286(3): pp. 616-21 , incorporated herein by reference). GLUT4 is translocated to the surface of myocytes and adipocytes in response to insulin binding to its receptor. Various proteins control this GLUT4 translocation, including VAMP2 and syntaxin-4. VAPA interacts with both of these proteins in skeletal myoblasts, and is suggested to be a regulator of VAMP2 availability in insulin- dependent GLUT4 translocation (Foster, et al., Traffic, 2000, 1(6): pp. 512-21 , incorporated herein by reference). The effect of insulin on GLUT4 translocation in monocytes is discussed, but there are indications that systemic insulin resistance is indicated by the presence of GLUT4 receptors on the monocyte surface (Mavros et al., Diabetes Res Clin Pract, 2009, 84(2): pp. 123-31, incorporated herein by reference. There is a strong association between variation in the SNPs rs29095, rs7237794, and rs917688 (Figure 1A) and insulin resistance, modestly reflected in gene expression, showing that the subjects with decreased leukocyte expression of VAPA during the AHC diet experience an increased insulin resistance. This suggests that the chromosome region where these SNPs are located is a susceptibility locus concerning insulin resistance. It remains to be seen if leukocytes have a role as insulin target cells. The genetic variability in VAPA, eventually contributing to a change in insulin resistance, may be caused by stronger gene expression changes in cells traditionally regarded as insulin target cells. As far as we know, this is the first time an association is found between genetic variability in VAPA and insulin resistance. Earlier the SNP rs29066, located in the 3 ^' UTR region of VAPA, between rs917688 and rs29095 has been found associated with bipolar disorder (Lohoff et al., J Neural Transm, 2008, 1 15(9): pp. 1339-45, incorporated herein by reference).

There are not many known genes regulated by the transcription factor PIAS1 , but three of them, myogenin (MYOG) (Hsuet al, J Biol Chem, 2006, 281(44): pp. 33008-18, incorporated herein by reference), actin, alpha 2, smooth muscle, aorta (ACTA2, member of F-actin) (Kawai-Kowase et al., Mol Cell Biol, 2005, 25(18): pp. 8009-23, incorporated herein by reference), and cyclin-dependent kinase inhibitor 1A (CDKN1A) (Megidish et al., J Biol Chem, 2002, 277(10): pp. 8255-9, incorporated herein by reference) are all mediators of insulin induced signalling, shown in a variety of cells, including neutrophils, adipocytes, myocytes, pancreatic islet cells, and intestinal endocrine cells. (See Chodniewicz and Zhelev, Blood, 2003, 102(6): pp. 2251-8; Inoue et al., J Biol Chem, 2008, 283(30): pp. 21220-9; Kaneto et al., Diabetologia, 1999, 42(9): pp. 1093-7; Lim et al., Endocrinology, 2009, 150(12): pp. 5249-61 ; Sumitani et al., Endocrinology, 2002, 143(3): pp. 820-8; and Yoshizaki et al., Mol Cell Biol, 2007, 27(14): pp. 5172-83; all incorporated herein by reference.) The association we found between the SNP rs649471 1 and insulin resistance showed that homozygotes for both the consensus and the alternative allele had a decrease in insulin resistance during the BMC diet, but the heterozygotes had no significant change. However, the genotype specific change was not reflected in the mRNA expression data of PIAS1, but the effect of the transcription factors could be controlled by post-transcriptional activation. The effect may also be mediated through gene expression responses in other cells more insulin sensitive than leukocytes.

Increased PAI-1 concentration in the liver is associated with insulin resistance in mice (Takeshita et al., Metabolism, 2006, 55(1 1): pp. 1464-72, incorporated herein by reference), and loss of affinity between GIP and GIP-receptor affect localization of PAI-1 to mouse plasma (Hansotia et al., J Clin Invest, 2007, 1 17(1): pp. 143-52, incorporated herein by reference). Since GIP-secretion is stimulated by glucose, this could explain why genetic variation in the GIP gene was associated with changes in PAI-1 protein concentrations in plasma.

Today the recommendation of daily intake of carbohydrates in Norway is 50-60 E% (Utviklingen i norsk kosthold, Vol. 2008, Utviklingen i norsk kosthold 2008, Oslo: Direktoratet, 2008, 27 s, incorporated herein by reference). Such a high fraction will contribute to a high dietary glycemic load, unless considerable caution is taken to choose carbohydrate sources with low glycemic index. With precaution, regarding the small sample size, our results suggest that some individuals are sensitive to high glycemic load, which is shown by an increase in insulin resistance during high-carbohydrate dieting (AHC) (Figure 2A). The same individuals have a distinct genotype profile for the SNPs most highly associated with changes in insulin resistance. Likewise, there are subjects that benefit more than others from low dietary glycemic load (Figure 2B), also with a distinct genotype profile. The observation that a significant number of these SNPs are located in genes already associated with T2D and other traits related to insulin resistance strengthens our hypothesis that one could discern strong and weak responders to glycemic load, by their genotype profile. However, our contribution to identify these QTLs affecting insulin resistance should be corroborated in larger studies. Reliable personalized nutritional advice is something still far ahead, and the theme may also raise considerable ethical debate, but our results suggest that the population at large, but especially subjects predisposed to develop T2D, should be aware of the glycemic challenge that a diet with high glycemic load gives. The use of genotyping data to link gene expression differences with phenotypes has increased markedly the last years. However, the use of genetic variation to stratify responses to a homeostatic challenge, like a diet intervention, has not been quite as common. The reason might be that the sample size required to gain significant results far exceeds what is easily manageable in an intervention study. We have shown that genotype is a source of interindividual variability in the response to a change in glycemic load, and suggest that genotype information can be integrated as an explanatory variable in microarray gene expression analysis.

Some obvious limitations need to be acknowledged in our study. What is already mentioned is the limited sample size. Whereas the study of average responses to a dietary intervention in a controlled cross-over study has produced robust findings (Arbo I, Brattbakk HR, Langaas M et al., A balanced macronutrient diet induces changes in a host of proinflammatory biomarkers, rendering a more healthy phenotype; a randomized cross-over trial, 2010, {manuscript submitted), incorporated herein by reference), dividing the subjects into two or three groups based on genotype inevitably decreases the statistical power. We nevertheless used reasonable criteria to declare associations of SNPs and eQTL (FDR < 0.05), while acknowledging that of course in larges studies much more significant association confidence can be obtained. We considered various quality criteria of the SNPs that could account for aberrant behaviour in our statistical tests. One quality criterion concerns the Hardy-Weinberg equilibrium (HWE). In a population, deviation from HWE may be indicative of selective pressure, but because most genes are not under selection it can also be used as an indicator of problems in the genotyping procedure leading to bias in the observed allele frequencies (Greene et al., Lect Notes Comput Sci, 2010, 6023(LNCS): pp. 74-85, incorporated herein by reference). Another reason why the allele distribution might deviate from HWE is the relatively small sample size of the current study, making it vulnerable to biased selection of subjects. Nevertheless, except where noted, none of the SNPs for which we found associations showed a gross skewness in allele frequencies that would significantly violate HWE, and indeed all passed the SNP call quality criteria of the Genotyping VI .1.9 software (Illumina). To ascertain the genotyping and HWE quality of each individual SNP is challenging, so we did carefully consider these quality criteria when interpreting the results of individual SNPs. Hierarchical clustering and PCA revealed two genetical outliers in our sample size (Figure 3). Why these subjects deviate from the others is not known, but to re-analyse the data without these outliers would be a reasonable approach.

Leukocytes are an easy accessible source for transcriptome profiling, and an obvious choice to screen for inflammatory gene expression changes in response to food. However, the knowledge on insulin responsiveness is limited. The inflammatory properties of monocytes and macrophages are central in the development of insulin resistance in insulin target cells, like adipocytes and myocytes. But it is not known whether the established molecular mechanisms behind insulin resistance are the same in leukocytes. We have shown earlier that monocytes are insulin responsive in a dose dependent manner (Ingerid Arbo, Cathinka L Halle, Darshan Malik, et al. Insulin induces fatty acid desaturase expression in human monocytes, 2010, (manuscript submitted), incorporated herein by reference), inducing increased desaturase transcription. However, this does not guarantee that we can expect significant association between leukocyte gene expression and changes in insulin resistance, considering an earlier finding that gene expression profiles in leukocytes and adipocytes deviate (Brattbakk HR, Arbo I, Aagaard S, et al. Balanced caloric macronutrient composition downregulates immunological gene expression in human blood cells - adipose tissue diverges, 2010, (manuscript submitted), incorporated herein by reference). This demonstrates the need to investigate not only blood, but also additional parallel sampled biopsies of well established insulin target tissue, like adipose tissue.

Of the SNPs disclosed herein, it is seen that some are directly related to the genes for VAPA, Pias l and GIP, while some are closely related thereto and can serve as "surrogate" markers. These SNPs are more specifically: rsl6961756, rsl242483, rs29095, rs7237794, rs917688, rs649471 1, rsl489595 and rs2291726.

The SNPs may serve as new markers of candidate QTL contributing to explain the genetic aspect of insulin resistance development. Also, VAPA and PIAS1 are new candidate genes involved in the molecular mechanisms behind insulin resistance. Finally, certain SNPs are candidate eQTL for plasma PAI-1 concentration, also related to insulin resistance. Our results have demonstrated the added value of incorporating genotype data in gene expression analysis to explain interindividual variability. A genotype profile of specific SNPs can distinguish weak and strong responders to glycemic load, with respect to insulin resistance. SNP typing may eventually be used to provide concrete dietary advice to persons genetically predisposed to T2D.

References

1. Tripathy, D. and A.O. Chavez, Defects in insulin secretion and action in the pathogenesis of type 2 diabetes mellitus. Curr Diab Rep, 2010. 10(3): p. 184-91.

2. Pagel-Langenickel, I., et al., The role of mitochondria in the pathophysiology of skeletal muscle insulin resistance. Endocr Rev, 2010. 31 (1): p. 25-51.

3. Facchini, F.S., et al., Insulin resistance as a predictor of age-related diseases. J Clin Endocrinol Metab, 2001. 86(8): p. 3574-8.

4. McLaughlin, T., et al., Prevalence of insulin resistance and associated cardiovascular disease risk factors among normal weight, overweight, and obese individuals. Metabolism, 2004. 53(4): p. 495-9.

5. Meigs, J.B., Epidemiology of type 2 diabetes and cardiovascular disease: translation from population to prevention: the Kelly West award lecture 2009. Diabetes Care, 2010. 33(8): p. 1865-71.

6. Ludwig, D.S., The glycemic index: physiological mechanisms relating to obesity, diabetes, and cardiovascular disease. Jama, 2002. 287(18): p. 2414-23.

7. Lairon, D., et al., Nutrigenetics: links between genetic background and response to Mediterranean-type diets. Public Health Nutr, 2009. 12(9A): p. 1601 -6.

8. Kaput, J., Nutrigenomics research for personalized nutrition and medicine. Curr Opin Biotechnol, 2008. 19(2): p. 1 10-20.

9. Martinez, J.A., et al., Genotype-dependent response to energy-restricted diets in obese subjects: towards personalized nutrition. Asia Pac J Clin Nutr, 2008. 17 Suppl 1 : p. 1 19-22.

10. Goldberg, R.B., Cytokine and cytokine-like inflammation markers, endothelial dysfunction, and imbalanced coagulation in development of diabetes and its complications. J Clin Endocrinol Metab, 2009. 94(9): p. 3171-82.

1 1. Olefsky, J.M. and C.K. Glass, Macrophages, inflammation, and insulin resistance. Annu Rev Physiol, 2010. 72: p. 219-46.

12. Hotamisligil, G.S., Inflammation and metabolic disorders. Nature, 2006. 444(7121): p. 860-7. 13. Kushner, I., D. Samols, and M. Magrey, A unifying biologic explanation for "high- sensitivity" C-reactive protein and "low-grade" inflammation. Arthritis Care Res (Hoboken), 2010. 62(4): p. 442-6.

14. O'Keefe, J.H., N.M. Gheewala, and J.O. O'Keefe, Dietary strategies for improving post-prandial glucose, lipids, inflammation, and cardiovascular health. J Am Coll Cardiol, 2008. 51(3): p. 249-55.

15. Barclay, A.W., et al., Glycemic index, glycemic load, and chronic disease risk~a meta-analysis of observational studies. Am J Clin Nutr, 2008. 87(3): p. 627-37.

16. McKeown, N.M., et al., Carbohydrate nutrition, insulin resistance, and the prevalence of the metabolic syndrome in the Framingham Offspring Cohort. Diabetes Care, 2004. 27(2): p. 538-46.

17. Qi, L. and F.B. Hu, Dietary glycemic load, whole grains, and systemic inflammation in diabetes: the epidemiological evidence. Curr Opin Lipidol, 2007. 18(1): p. 3-8.

18. Voight, B.F., et al., Twelve type 2 diabetes susceptibility loci identified through large- scale association analysis. Nat Genet, 2010. 42(7): p. 579-89.

19. Kantartzis, K., et al., The DGAT2 gene is a candidate for the dissociation between fatty liver and insulin resistance in humans. Clin Sci (Lond), 2009. 1 16(6): p. 531-7.

20. Liu, P.H., et al., Genetic variants of TCF7L2 are associated with insulin resistance and related metabolic phenotypes in Taiwanese adolescents and Caucasian young adults. J Clin Endocrinol Metab, 2009. 94(9): p. 3575-82.

21. Palmer, N.D., et al., Association of protein tyrosine phosphatase IB gene polymorphisms with measures of glucose homeostasis in Hispanic Americans: the insulin resistance atherosclerosis study (IRAS) family study. Diabetes, 2004. 53(1 1): p. 3013-9.

22. Richardson, D.K., et al., Association between variants in the genes for adiponectin and its receptors with insulin resistance syndrome (IRS)-related phenotypes in Mexican Americans. Diabetologia, 2006. 49(10): p. 2317-28.

23. Ruchat, S.M., et al., Associations between glucose tolerance, insulin sensitivity and insulin secretion phenotypes and polymorphisms in adiponectin and adiponectin receptor genes in the Quebec Family Study. Diabet Med, 2008. 25(4): p. 400-6.

24. Smith, S.R., et al., A promoter genotype and oxidative stress potentially link resistin to human insulin resistance. Diabetes, 2003. 52(7): p. 161 1-8. 25. Jansen, R.C. and J.P. Nap, Genetical genomics: the added value from segregation. Trends Genet, 2001. 17(7): p. 388-91.

26. Schadt, E.E., et al., Genetics of gene expression surveyed in maize, mouse and man. Nature, 2003. 422(6929): p. 297-302.

27. Rockman, M.V. and L. Kruglyak, Genetics of global gene expression. Nat Rev Genet, 2006. 7(1 1): p. 862-72.

28. Wallace, T.M., J.C. Levy, and D.R. Matthews, Use and abuse of HOMA modeling. Diabetes Care, 2004. 27(6): p. 1487-95.

29. Matthews, D.R., et al., Homeostasis model assessment: insulin resistance and beta- cell function from fasting plasma glucose and insulin concentrations in man. Diabetologia, 1985. 28(7): p. 412-9.

30. Levy, J.C, D.R. Matthews, and M.P. Hermans, Correct homeostasis model assessment (HOMA) evaluation uses the computer program. Diabetes Care, 1998. 21 (12): p. 2191-2.

31. Mojiminiyi, O.A. and N.A. Abdella, Effect of homeostasis model assessment computational method on the definition and associations of insulin resistance. Clin Chem Lab Med, 2010.

32. Illumina, GenomeStudioTM Genotyping Module vl. O User Guide. 2008, Illumina, Inc: San Diego, CA.

33. Wymann, M.P. and R. Schneiter, Lipid signalling in disease. Nat Rev Mol Cell Biol, 2008. 9(2): p. 162-76.

34. Yuan, H.Y., et al., FASTSNP: an always up-to-date and extendable service for SNP function analysis and prioritization. Nucleic Acids Res, 2006. 34(Web Server issue): p. W635-41.

35. Benjamini, Y. and Y. Hochberg, Controlling the False Discovery Rate: A Practical and Powerful Approach to Multiple Testing. Journal of the Royal Statistical Society. Series B (Methodological), 1995. 57(1): p. 289-300.

36. R Development Core Team, R: A Language and Environment for Statistical Computing. 2010; Available from: www.R-project.org.

37. Povel, CM., et al., Glucose levels and genetic variants across transcriptional pathways: interaction effects with BMI. Int J Obes (Lond), 2010. 34(5): p. 840-5. 38. Weir, M.L., et al., VAP-A binds promiscuously to both v- and tSNAREs. Biochem Biophys Res Commun, 2001. 286(3): p. 616-21.

39. Foster, L.J., et al., A functional role for VAP-33 in insulin-stimulated GLUT4 traffic. Traffic, 2000. 1 (6): p. 512-21.

40. Mavros, Y., D. Simar, and M.A. Singh, Glucose Tranporter-4 expression in monocytes: a systematic review. Diabetes Res Clin Pract, 2009. 84(2): p. 123-31.

41. Lohoff, F.W., et al., Association between polymorphisms in the vesicle-associated membrane protein-associated protein A (VAPA) gene on chromosome 18p and bipolar disorder. J Neural Transm, 2008. 1 15(9): p. 1339-45.

42. Hsu, Y.H., et al., Sumoylated SnoN represses transcription in a promoter-specific manner. J Biol Chem, 2006. 281(44): p. 33008-18.

43. Kawai-Kowase, K., et al., PIAS 1 activates the expression of smooth muscle cell differentiation marker genes by interacting with serum response factor and class I basic helix- loop-helix proteins. Mol Cell Bio), 2005. 25(18): p. 8009-23.

44. Megidish, T., J.H. Xu, and C.W. Xu, Activation of p53 by protein inhibitor of activated Stat] (PIAS1). J Biol Chem, 2002. 277(10): p. 8255-9.

45. Chodniewicz, D. and D.V. Zhelev, Novel pathways of F-actin polymerization in the human neutrophil. Blood, 2003. 102(6): p. 2251-8.

46. Inoue, N., et al., Cyclin-dependent kinase inhibitor, p21 WAF1/CIP1, is involved in adipocyte differentiation and hypertrophy, linking to obesity, and insulin resistance. J Biol Chem, 2008. 283(30): p. 21220-9.

47. Kaneto, H., et al., Oxidative stress induces p21 expression in pancreatic islet cells: possible implication in beta-cell dysfunction. Diabetologia, 1999. 42(9): p. 1093-7.

48. Lim, G.E., et al., The rho guanosine 5'-triphosphatase, cell division cycle 42, is required for insulin-induced actin remodeling and glucagon-like peptide-1 secretion in the intestinal endocrine L cell. Endocrinology, 2009. 150(12): p. 5249-61.

49. Sumitani, S., et al., Aktl and Akt2 differently regulate muscle creatine kinase and myogenin gene transcription in insulin-induced differentiation of C2C12 myoblasts. Endocrinology, 2002. 143(3): p. 820-8.

50. Yoshizaki, T., et al., Myosin 5a is an insulin-stimulated Akt2 (protein kinase Bbeta) substrate modulating GLUT4 vesicle translocation. Mol Cell Biol, 2007. 27(14): p. 5172-83. 51. Takeshita, Y., et al., Tumor necrosis factor-alpha-induced production of plasminogen activator inhibitor 1 and its regulation by pioglitazone and cerivastatin in a nonmalignant human hepatocyte cell line. Metabolism, 2006. 55(1 1 ): p. 1464-72.

52. Hansotia, T., et al., Extrapancreatic incretin receptors modulate glucose homeostasis, body weight, and energy expenditure. J Clin Invest, 2007. 117(1): p. 143-52.

53. Utviklingen i norsk kosthold. Vol. 2008, Utviklingen i norsk kosthold 2008., Oslo: Direktoratet, 2008. 27 s.

54. Greene, C.S., D.S. Himmelstein, and J.H. Moore, A Model Free Method to Generate Human Genetics Datasets with Complex Gene-Disease Relationships. Lect Notes Comput Sci, 2010. 6023(LNCS): p. 74-85.

All references cited herein, are hereby incorporated by reference. Sequences of the single nucleotide polymorphisms cited by the accession numbers herein, are hereby incorporated by reference and can be found at www.ncbi.nlm.nih.gov/sites/entrez or snpper.chip.org/bio, among other sites, using the accession numbers provided.

TABLES

Table 1 Biological functions and diseases related to the SNPs that showed highest association with HOMA2 IR. The genes linked to the SNPs in the ref-SNP selection Top 100 lists were compared with the genes linked with the SNPs on the Metabochip. Significantly enriched IPA defined functions and diseases, according IPA's Functional Analysis, are listed (FDR < 0.01 ). The table also shows the number of genes related to the functions, linked with the SNPs in the ref-SNP selection Top 100 lists for the AHC6-AHC0 and the BMC6-BMC0 comparisons separately.

All Top 100 lists AHC BMC

Function Annotation P-value # genes # genes # genes

diabetes mellitus 5.17E-13 63 15 23

T2D 6.39E-13 47 10 20

endocrine system disorder 1.68E-12 64 - 24

metabolic disorder 3.48E-12 67 16 24

cardiovascular disorder 6.46E-10 58 20 17

hypertension 8.65E-09 35 12 - atherosclerosis 4.56E-08 38 13 - genetic disorder 4.56E-08 102 27 35

coronary artery disease 5.01E-08 36 12 12

T1 D 1.80E-07 33 1 1 - autoimmune disease 8.59E-07 50 17 16

immunological disorder 1.96E-06 54 18 17

amyotrophic lateral sclerosis 1.53E-05 21 6 10

progressive motor neuropathy 1.68E-05 31 1 1 -

Crohn's disease 5.41E-05 29 - 1 1

neurological disorder 6.55E-05- 66 19 26

rheumatoid arthritis 1.50E-04 35 13 10

digestive system disorder 2.48E-04 33 - 12

inflammatory disorder 5.16E-04 53 17 -

Alzheimer's disease 9.78E-04 23 6 10

skeletal and muscular disorder 1.80E-03 51 18 15

rheumatic disease 1.93E-03 36 _ connective tissue disorder 2.23E-03 Parkinson's disease 5.24E-03

Table 2 Pairs of associations between SNP and the gene expression values in nearest gene in the gene-SNP selection Top 100 lists. Pairs in which are included, the gene is related to insulin resistance/T2D in at least one PubMed entry, and there is at least one significant (FDR < 0.05, value typed in bold) genotype specific gene expression change (log ₂ -ratio) for one of the four comparisons, and a GenTrainScore > 0.750.

PDZD2 rs283122 C ~> T 0.099 0.866 o

u CC 1 NA NA

X

< CT 1 1 0.059 0.036

u TT 18 -0.031 0.086

<

TPM1 1 rs 17752921 T --> C 0.291 0.850

TT 27 -0.123 < 0.001

CT 3 0.122 0.378

LPA 9 rs6415084 T ~> c 0.055 0.838

CC 6 0.016 0.801

CT 18 0.019 0.307

TT 6 -0.127 0.039

NME1 1 rs2302254 C ~> T 0.234 0.769

CC 23 -0.058 0.108

CT 7 0.126 0.042

ADRA 1A 1 rs4732874 T ~ c 0.462 0.879

CC 17 -0.070 0.031

CT 10 0.070 0.108

TT 3 -0.027 0.431

ARHGEF1 1 3 rs822570 T ~> c 0.030 0.927

CC 1 1 -0.061 0.417 CT 12 0.229 0.005

TT 7 0.012 0.910

S1K3 1 rs888246 C ~> T 0.463 0.838

CC 25 -0.084 0.130

CT 5 -0.429 0.031

TMEM195 1 rs7781413 A --> G 0.943 0.869

AA 21 -0.036 0.021

AG 8 0.057 0.280

GG 1 NA NA

ADCY9 1 rs2532007 G --> A 0.424 0.856

AA 24 0.038 0.312

AG 6 -0.226 0.184

SLC2A1 6 rs751210 G --> A 0.051 0.808

AA 1 NA NA

AG 14 -0.1 10 0.328

GG 15 -0.513 0.002

PLTP 9 rs435306 T ~> G 0.495 0.847

AA 19 -0.039 0.048

AC 9 0.047 0.180

CC 2 NA NA

rs378114 C -> 0.495 0.904

AA 2 NA NA

AG 9 0.047 0.180

GG 19 -0.039 0.048

CAMTA1 1 rs6577435 C — > A 0.440 0.873 o

AA 1 NA NA

AC - 10 0.105 0.232

u CC 21 -0.1 17 0.009

CACNAIG 1 rs989128 G ~> A 0.012 0.785

AA 6 -0.017 0.513

AG 10 -0.081 0.028

GG 16 0.013 0.323 SULF2 1 rs6125103 C ~> T 0.203 0.837 cc 22 0.210 0.011

CT 9 -0.128 0.220

TT 1 NA NA

ITGAV 1 rs3738919 C --> A 0.595 0.925

AA 8 0.515 0.032

AC 14 -0.018 0.764

CC 10 -0.060 0.657

SULF2 1 rsl 1699888 C --> T 0.421 0.862

CC 27 -0.054 0.294

CT 5 0.359 0.039

FIGURE LEGENDS

Figure 1 Changes in HOMA2 IR or protein concentration (log ₂ -ratio), separately for each comparison, and genotype for the SNPs indicated. GenTrain score > 0.73 for all SNPs (A-D). The SNP rs2291726 (D) deviated from HWE (P < 0.05).

Figure 2 Unsupervised hierarchical clustering (Manhattan distance measures, complete linkage) of the Top 100 SNPs (rows) associated with change in HOMA2 IR in response to A) the AHC diet, and B) the BMC diet. The subjects (columns) are sorted by HOMA2 IR change, increasing from left to right. SNPs within the right hand side brackets of the heatmaps are identified in the ref-SNP selection Top 100 lists in Supplementary table 1-4. Figure 3 A) Hierarchical clustering showing distance between subjects based on genotype information from the 71 061 SNPs (Manhattan distance measures, complete linkage). B) PCA plot based on the same data, showing the 3 first principal components.

Supplementary Table 1 Ref-SNP selection ToplOO list, AHC6-AHC0

SNP Nearest Gene ANOVA p-value GenTrain Score ChiTestlOO Clusl rsl 6961756 <0.001 0.736 0.421 B rs29095 VAPA <0.001 0.875 0.657 B rs7237794 VAPA <0.001 0.843 0.688 B rs917688 <0.001 0.738 0.657 B rsl0803976 <0.001 0.885 0.463 A rs780242 O.001 0.904 0.091 A rs 16914660 ANK3 <0.001 0.842 0.667 A rs7600698 <0.001 0.507 0.362 rsl 7236914 <0.001 0.845 0.072 B rs6753302 EPAS1 <0.001 0.837 <0.001 B rsl 1858742 ZWILCH O.001 0.832 0.354 B rsl 1031821 <0.001 0.901 0.152 A rs6445062 <0.001 0.877 0.067 rs4646400 PEMT <0.001 0.798 0.463 A rs3828760 FAM46A <0.001 0.885 0.004 B rsl390785 GALNTL6 <0.001 0.901 0.742 A rsl0509771 CCDC147 <0.001 0.891 0.004 A rs2480 <0.001 0.912 0.51 1 rs968371 CSMD1 <0.001 0.81 1 0.670 A rsl 3176923 <0.001 0.751 0.657 B rs l 3189446 <0.001 0.824 0.688 B rs7243663 L3MBTL4 <0.001 0.771 0.688 B rsl 7501809 C9orf46 <0.001 0.804 0.352 B rsl 1038913 AMBRAl <0.001 0.841 0.163 B rsl7401 147 <0.001 0.885 0.073 B rs42495 SEMA5A O.001 0.915 0.857 rsl 0972856 <0.001 0.777 0.234 B rsl 15701 15 MYBPC3 <0.001 0.888 0.144 B rs4579523 <0.001 0.802 O.001 rs7101470 C l lorf49 <0.001 0.917 0.655 B rsl 0838651 Cl lorf49 <0.001 0.903 0.655 A rsl6843307 <0.001 0.936 0.073 B rsl2666730 <0.001 0.897 0.857 rs236231 1 ABCA13 <0.001 0.826 0.072 A rsl 1 179215 TRHDE <0.001 0.898 0.295 B rsl 883414 LOCI 00294320 0.001 0.929 0.063 A rsl0121339 0.001 0.819 0.421 B rs41431 10 0.001 0.935 0.424 B rsl2973523 FCER2 0.001 0.771 0.425 rsl2486603 MYRIP 0.001 0.914 0.01 1 A rs7071851 PTCHD3 0.001 0.893 0.523 rs2817644 0.001 0.816 0.574 B rs6902530 0.001 0.784 0.657 B rsl0483213 CENPM 0.001 0.767 0.495 B rsl 1700328 ANGPT4 0.001 0.784 0.425 rsl6824470 0.001 0.91 1 0.863 rs81581 1 C2orf61 0.001 0.795 0.144 rs2350623 C22orf9 0.001 0.889 0.285 rsl 0242065 0.001 0.830 0.018 rsl 3227663 0.001 0.860 0.027 rs7836548 0.001 0.864 0.109 B rsl 1790360 0.001 0.81 1 0.474 rs2262933 SMYD3 0.001 0.863 0.526 rs685897 0.001 0.835 0.846 A rs4239424 0.001 0.712 0.001 rsl409570 0.001 0.695 0.185 rs7897931 0.001 0.836 0.339 rsl 1022039 0.001 0.873 0.523 rs7553849 PRDM16 0.001 0.750 0.079 A rs2569430 CTU1 0.001 0.778 0.015 rs4994351 ZNF331 0.001 0.741 0.049 rsl2366082 DSCAMLl 0.001 0.746 0.290 B rsl91861 1 ABCA13 0.001 0.872 0.268 A rs6871607 0.001 0.865 0.574 B rsl 1649247 WWOX 0.001 0.707 0.462 A rsl 7722827 0.001 0.754 0.421 B rsl 7766326 SLC25A21 0.001 0.863 0.495 B rs5999900 0.001 0.713 0.354 B rsl 885750 0.001 0.724 0.162 A rsl2504564 TMEM144 0.001 0.826 0.058 rsl 82390 0.001 0.896 0.495 A rsl0153481 ZNF709 0.001 0.682 0.002 rsl 7706149 NUP35 0.001 0.876 0.290 B rs7138803 0.001 0.854 0.041 A rsl0928303 0.001 0.858 0.291 B rsl437848 GALNTL6 0.001 0.907 0.354 B rs2010014 0.001 0.879 0.352 B rs221020 PAK7 0.001 0.915 0.667 A rsl0916785 0.001 0.741 0.001 rs270413 BMP6 0.001 0.826 0.146 A rsl297215 NR1P1 0.001 0.910 0.064 A rs2253231 NRIP1 0.001 0.912 0.064 A rs7729395 0.001 0.876 0.185 rs2242104 VLDLR 0.001 0.894 0.189 A rs767145 0.001 0.808 0.835 A rs739571 0.001 0.901 0.336 A rsl 7668126 0.001 0.797 0.336 rs2432761 FARS2 0.002 0.913 0.672 rs3751544 MEIS2 0.002 0.885 0.425 A rs6663310 0.002 0.756 0.001 rs431 1480 FILIP1 0.002 0.890 0.830 rs2241733 PLXNA4 0.002 0.763 0.234 B rs2160706 0.002 0.836 0.225 rsl 1 129948 0.002 0.827 0.600 A rs6859734 ADAMTS 19 0.002 0.863 <0.001 A rs7380441 ADAMTS 19 0.002 0.805 O.001 rsl 319075 0.002 0.820 0.057 A rs73821 12 0.002 0.906 0.019 A rs9393921 0.002 0.880 0.021 rs9885928 0.002 0.820 0.057

Supplementary Table 2 Ref-SNP selection ToplOO list, BMC6-BMC0

ANOVA p- GenTrain

SNP Nearest Gene value Score ChiTestlOO Cluster rs2623763 <0.001 0.670 0.088

rs4712572 CDKAL1 <0.001 0.760 0.062

rsl 993919 STAB2 <0.001 0.881 0.144 D rs7536825 KIF26B <0.001 0.846 0.131 C rsl 6960303 CDH13 <0.001 0.903 0.142

rs9902569 <0.001 0.922 0.667 E rsl 3244124 CDK14 <0.001 0.819 0.018 D rs6052937 SLC23A2 <0.001 0.884 O.001 E rsl 2359453 <0.001 0.789 0.742 D rs6069099 <0.001 0.837 0.285

rsl 7620466 <0.001 0.738 0.495 D rs9961435 <0.001 0.788 0.162 E rs875294 <0.001 0.815 0.234 C rs4465666 <0.001 0.606 <0.001 C rsl 1912637 <0.001 0.883 0.268

rsl 0431808 CLN6 <0.001 0.822 0.888

rs649471 1 PIAS1 <0.001 0.924 0.888 C rs4895769 <0.001 0.887 0.225 E rsl 1855184 DMXL2 <0.001 0.897 0.049 E rsl 0512215 <0.001 0.899 0.526

rsl 3122545 FAM190A <0.001 0.892 0.291 D rs9466015 <0.001 0.917 0.742 D rs955010 FAM190A <0.001 0.916 0.290 D rs4704320 1QGAP2 <0.001 0.877 0.399 E rs3807689 MAGI2 <0.001 0.818 0.672 D rs4650540 <0.001 0.930 0.260

rsl 3356198 CHSY3 <0.001 0.875 0.440

rs6463750 <0.001 0.855 0.027

rs257215 <0.001 0.875 0.009 rs257221 <0.001 0.858 0.009 rs3746532 LOC100287002 0.001 0.673 0.508 rsl 1107935 NAV3 0.001 0.916 0.352 D rsl 359292 CCDC30 0.001 0.737 0.290 D rsl 0896450 0.001 0.838 0.595 rs4620729 0.001 0.802 0.3 18 E rs7947353 0.001 0.902 0.595 E rs6706382 0.001 0.877 0.244 rsl7047703 TGFB2 0.001 0.801 0.871 rsl 1581605 TGFB2 0.001 0.924 0.871 rs7950547 0.001 0.840 0.828 rsl0793139 0.001 0.824 0.799 rsl2742404 CAMSAPILI 0.001 0.951 0.073 D rs2292096 CAMSAPILI 0.001 0.902 0.073 D rs2514801 CDH17 0.001 0.875 0.349 rs2338545 PLB1 0.001 0.810 0.440 rsl 1675205 TCF7L1 0.001 0.918 0.614 rs4710944 CDKAL1 0.001 0.884 0.244

rs627522 ZNF708 0.001 0.894 0.724

rs4774183 0.001 0.507 O.OOl rsl2065336 0.001 0.728 0.290 D rs950692 GPR98 0.001 0.751 0.234 D rsl 1712666 VGLL4 0.001 0.733 0.131 rsl 2546518 GRHL2 0.001 0.808 0.152

rs3091317 0.001 0.903 0.899

rs3091321 CCL7 0.001 0.901 0.863 E rsl2891473 SRP54 0.001 0.698 0.001 rs6989246 MY0M2 0.001 0.815 0.943 E rsl l 871821 0.001 0.685 <0.001 rsl7746008 PHLPP1 0.001 0.895 0.502 D rsl 799977 MLH1 0.001 0.936 0.614

rs807013 0.001 0.693 0.003 rs 1907415 0.001 0.935 0.526 rs7616047 0.001 0.886 0.146 rsl0735653 0.001 0.829 <0.001 E rs2590174 0.001 0.861 0.924 C rs35879596 GRAMDIA 0.001 0.887 0.440 E rs4290308 0.001 0.822 0.042 C rs 1432226 THSD7B 0.001 0.804 0.075 rs 1 1042902 MR VII 0.001 0.872 0.459 rs979015 0.001 0.807 0.137 C rs4555526 0.001 0.910 0.017 rs2425463 CHD6 0.001 0.821 0.036 rsl 7637580 LARS2 0.001 0.822 0.015 E rsl 0465729 0.001 0.833 0.01 1 rs 10038804 UGT3A2 0.001 0.865 0.320 rs2038431 ZFP64 0.001 0.726 0.943 rs7604914 FAM82A1 0.001 0.898 0.916 rs3900452 0.001 0.762 0.195 rs4016189 0.001 0.890 0.195 C rs2159894 0.001 0.773 0.672

rsl 7674590 0.001 0.81 1 0.295 E rsl 156619 0.001 0.912 0.244

rs922453 0.001 0.882 0.667 E rs654126 CSMD1 0.001 0.909 0.108 rs6927578 PARK2 0.001 0.815 0.463 D rsl l 950170 0.001 0.849 0.421 D rsl 6823728 C2orf83 0.001 0.776 0.688 D rsl 7633078 KATNAL1 0.001 0.883 0.502 D rs277315 0.001 0.736 0.502 D rs9562933 0.001 0.864 0.042

rs716453 PPAPDC 1 A 0.001 0.731 0.657 D rs7686154 0.001 0.865 0.023 D rs7968178 0.001 0.91 1 0.657 D rs226236 LASP1 0.001 0.804 0.177 rs2943599 0.001 0.869 0.318 rsl0853522 0.001 0.906 0.924 E rs 192671 CCDC50 0.001 0.881 0.441 E rs6502774 TUSC5 0.001 0.808 0.268 rs4905899 EML1 0.001 0.788 0.001 rs2028210 AMPH 0.001 0.890 0.672

Supplementary Table 3 Ref-SNP selection Top 100 list, BMC6-AHC6

Nearest ANOVA p- GenTrain

SNP Gene value Score ChiTestlOO rsl2304001 <0.001 0.826 0.399 rsl7236914 <0.001 0.845 0.072 rs6753302 EPAS 1 <0.001 0.837 <0.001 rsl6916966 <0.001 0.878 0.001 rsl 556260 USF1 <0.001 0.918 0.005 rs7597683 <0.001 0.814 0.094 rs7160372 <0.001 0.804 0.267 rs7196505 <0.001 0.849 0.225 rs l996806 RGS7 <0.001 0.881 0.462 rs l l 558471 SLC30A8 <0.001 0.906 0.846 rs9296579 <0.001 0.808 <0.001 rs 12468863 KCNK3 <0.001 0.834 0.295 rs 1275941 <0.001 0.855 0.549 rs3739081 <0.001 0.940 0.549 rs6859734 ADAMTS19 <0.001 0.863 <0.001 fs7380441 ADAMTS19 <0.001 0.805 <0.001 rs 10220965 <0.001 0.756 0.109 rsl488666 <0.001 0.693 0.506 rs2781792 <0.001 0.741 0.185 rs 17069214 <0.001 0.931 0.203 rsl 7245857 <0.001 0.827 0.055 rs7553849 PRDM16 <0.001 0.750 0.079 rs2235642 IFT140 <0.001 0.694 0.667 rs3758376 SEC61A2 <0.001 0.836 0.1 16 rs2305413 CHRNA 1 <0.001 0.909 0.424 rsl2903587 CHD2 <0.001 0.853 0.393 rs2062096 <0.001 0.938 <0.001 rs 12467466 CENPA <0.001 0.754 0.362 rs3802177 SLC30A8 <0.001 0.932 0.672 rsl 1 179215 TRHDE O.001 0.898 0.295 rs2399786 NUDT5 <0.001 0.910 0.659 rs6744164 0.001 0.812 0.421 rs968371 CSMD1 0.001 0.81 1 0.670 rsl 3266634 SLC30A8 0.001 0.881 0.914 rs7855478 M0RN5 0.001 0.692 0.109 rsl 2424799 0.001 0.773 0.080 rsl 2891948 0.001 0.754 0.320 rsl 7801467 0.001 0.840 0.290 rs929269 ENDOU 0.001 0.727 0.421 rs281385 MAMSTR 0.001 0.721 0.290 rsl2964419 0.001 0.919 0.871 rs2274305 DCDC2 0.001 0.862 0.290 rs488078 0.001 0.869 0.548 rsl 0947465 0.001 0.805 0.393 rsl 7484283 0.001 0.785 0.001 rsl2982980 ZNF468 0.001 0.638 0.022 rs4466385 C8orf34 0.001 0.881 0.067 rsl 7736747 0.001 0.849 0.295 rs4644227 C8orl34 0.001 0.863 0.51 1 rsl 7635121 0:001 0.934 0.019 rsl 674091 DTX1 0.001 0.689 0.080 rs472972 POLN 0.001 0.825 0.421 rsl487775 0.001 0.862 0.295 rs2072844 0.001 0.872 0.062 rsl 861699 0.001 0.874 0.587 rs3788464 SY 3 0.001 0.873 0.657 rs942024 0.001 0.752 0.672 rsl 2973523 FCER2 0.001 0.771 0.425 rsl 1964281 ESR1 0.001 0.771 0.421 rs7025024 0.001 0.795 0.549 rsl 2683791 0.001 0.936 0.030 rs7862653 0.001 0.779 0.349 rs870535 0.001 0.912 0.319 rs7944972 OPCML 0.001 0.852 0.002 rs582669 PKHD1 0.001 0.927 0.637 rsl31 16006 0.001 0.867 0.614 rs 1424790 0.001 0.830 0.020 rs 1625560 0.001 0.898 0.614 rs3777102 NRG2 0.001 0.777 0.857 rs912377 0.001 0.807 0.891 rsl3220430 EYS 0.001 0.819 0.422 rsl363472 KIAA1024L 0.001 0.836 <0.001 rsl71895 0.001 0.934 0.041 rs9592493 PCDH9 0.001 0.829 0.586 rs468471 RCL1 0.001 0.350 0.554 rs2338871 LCP2 0.001 0.741 0.143 rs2830957 0.001 0.883 0.586 rsl7718358 0.001 0.754 0.574 rs9645497 0.001 0.815 0.688 rs6777976 OXNAD1 0.001 0.783 0.225 rs38478 0.001 0.847 0.320 rs763842 0.002 0.920 0.339 rsl2413154 RHOBTB l 0.002 0.567 0.050 rs6995157 0.002 0.574 0.388 rsl0163354 ABCC1 1 0.002 0.919 <0.001 rs2504927 SLC22A3 0.002 0.890 0.523 rsl 3424541 ZNF638 0.002 0.869 0.574 rs3176295 FGF17 0.002 0.751 0.574 rs 17479629 MICAL2 0.002 0.903 0.657 rs9815875 0.002 0.752 0.349 rs2807304 TLE4 0.002 0.801 0.422 rs 1 1772485 0.002 0.740 0.058 rs 17098621 0.002 0.822 0.755 rs2063777 0.002 0.797 0.672 rsl0830089 0.002 0.773 0.399 rs3766509 ACP6 0.002 0.846 0.778 rs7267327 0.002 0.856 0.639 rsl7150506 CSNK1G3 0.002 0.910 0.336 rs2112468 CSNK1G3 0.002 0.912 0.506 rs4546375 CSNK1G3 0.002 0.900 0.336

Supplementary Table 4 Ref-SNP selection Top 100 list, (BMC6-BMC0)-(AHC6-AHC0)

ANOVA p- GenTrain

SNP Nearest Gene value Score ChiTestlOO

rs2480 <0.001 0.912 0.51 1

rs 1704405 EHD4 <0.001 0.757 0.586

rs7071851 PTCHD3 <0.001 0.893 0.523

rs204925 LMOl <0.001 0.817 0.004

rsl 19161 12 ARHGEF3 <0.001 0.813 0.007

rsl 1041982 STK33 <0.001 0.882 0.058

rs7538377 PCNXL2 <0.001 0.788 0.778

rsl409570 <0.001 0.695 0.185

rsl7138899 ACACA <0.001 0.907 0.354

rs2302803 ACACA <0.001 0.879 0.354

rs993743 <0.001 0.853 0.080

rsl 1 187169 <0.001 0.810 0.586

rs4712572 CDKAL1 <0.001 0.760 0.062

rsl 3176923 <0.001 0.751 0.657

rsl 3189446 <0.001 0.824 0.688

rsl 0017447 <0.001 0.789 0.001

rsl 1 101387 ARHGAP22 <0.001 0.808 0.149

rsl l 213776 <0.001 0.800 0.037

rsl 502275 <0.001 0.727 0.424

rsl7095168 <0.001 0.743 0.421

rsl0833451 NELL1 0.001 0.729 <0.001

rs6047259 0.001 0.818 0.891

rs807013 0.001 0.693 0.003

rs9523880 0.001 0.943 0.079

rs968371 CSMD1 0.001 0.81 1 0.670

rsl 1783921 0.001 0.855 0.574

rs3104917 0.001 0.844 0.502

rs3887267 C9orf3 0.001 0.868 0.688

rsl 1695576 0.001 0.754 0.407 rs l 1 193140 S0RCS 1 0.001 0.691 0.163 rs3744589 ACACA 0.001 0.945 0.495 rs7729395 0.001 0.876 0.185 rs7660651 0.001 0.929 0.203 rs l 1070879 MAPK6 0.001 0.921 0.399 rsl6843307 0.001 0.936 0.073 rs758504 NFIC 0.001 0.795 0.502 rs7897931 0.001 0.836 0.339 rs4810347 0.001 0.825 0.506 rs l 159051 1 0.001 0.817 0.001 rsl 860904 0.001 0.856 0.01 1 rs7677806 0.001 0.882 0.01 1 rs2460968 SAMD12 0.001 0.805 0.079 rsl0093536 0.001 0.884 0.667 rsl 0803976 0.001 0.885 0.463 rsl0242065 0.001 0.830 0.018 rsl 3227663 0.001 0.860 0.027 rs6748854 0.001 0.807 0.01 1 rsl2666730 0.001 0.897 0.857 rs3828760 FAM46A 0.001 0.885 0.004 rs654126 CSMD1 0.001 0.909 0.108 rs7106565 0.001 0.879 0.891 rs4579523 0.001 0.802 <0.001 rsl 0933436 0.001 0.770 0.339 rsl 1693862 0.001 0.776 0.339 rs9558407 0.001 0.816 0.003 rsl 0463168 0.001 0.91 1 0.659 rs6502774 TUSC5 0.001 0.808 0.268 rsl 546208 0.001 0.908 0.421 rs3735444 MAGI2 0.001 0.806 0.285 rsl 721073 0.001 0.910 0.399 rs963080 0.001 0.913 0.399 rs381 1976 SLC04C1 0.001 0.771 0.943 rs6891076 0.001 0.887 0.943 rsl 0929308 HEATR7B1 0.001 0.934 0.595 rs353747 0.001 0.840 0.336 rsl2891473 SRP54 0.001 0.698 0.001 rs6951227 MAGI2 0.001 0.849 0.088 rs6663310 0.001 0.756 0.001 rs7127684 STK33 0.001 0.897 0.024 rsl 3324043 0.001 0.792 0.463 rsl 1638978 0.001 0.829 0.799 rsl0124300 0.001 0.784 0.042 rsl 6914660 ANK3 0.001 0.842 0.667 rsl 2492974 0.001 0.778 0.234 rsl6838912 0.001 0.760 0.234 rsl 3028683 CDKL4 0.001 0.793 0.424 rsl018966 CTN D2 0.001 0.924 0.012 rsl7318596 ATP5SL 0.001 0.608 <0.001 rs4674 BCKDHA 0.001 0.81 1 <0.001 rs31 18942 LPPR1 0.001 0.838 0.027 rs6548940 0.001 0.785 0.093 rs6753302 EPAS 1 0.001 0.837 O.OOl rs236004 0.001 0.825 0.006 rs2413923 SHC4 0.001 0.894 0.586 rsl 077481 1 0.001 0.847 0.094 rs828999 SLC25A24 0.001 0.934 0.828 rs4787016 A2BP1 0.001 0.789 0.093 rsl 7545182 0.001 0.770 0.039 rsl 7236914 0.001 0.845 0.072 rs7243663 L3MBTL4 0.001 0.771 0.688 rsl 2735509 0.001 0.875 0.463 rsl 2065336 0.001 0.728 0.290 rs9918378 0.001 0.785 0.185 rs4236002 CDKAL1 0.001 0.931 0.336 rs 12619647 SEPT2 0.001 0.819 0.914 rs7313017 LOC100130825 0.001 0.512 0.021 rs9644620 LOC100128993 0.001 0.920 0.672 rs2599547 0.002 0.902 0.137 rs974312 0.002 0.748 0.495 rs6573513 PPP2R5E 0.002 0.854 0.049

Supplementary Table 5 Gene-SNP selection Top 100 list, AHC6-AHC0

Nearest ANOVA p-

SNP Gene value GenTrainScore ChiTestlOO rs6802942 PPP2R3A <0.001 0.891 0.495 rs6513775 PTPRT <0.001 0.776 0.088 rs4767020 RPH3A <0.001 0.743 0.495 rs9863749 C3orf20 <0.001 0.854 0.421 rs6035839 XRN2 <0.001 0.891 0.422 rs6082384 XRN2 <0.001 0.937 0.422 rs 10982661 TMOD1 <0.001 0.838 0.143 rs 10071707 PDZD2 <0.001 0.919 0.587 rs 17817463 DISCI <0.001 0.733 0.421 rs283122 PDZD2 <0.001 0.866 0.099 rs6959021 PKD1L1 <0.001 0.864 0.268 rs9639988 PKD1L1 <0.001 0.819 0.268 rs2345122 ZKSCAN2 <0.001 0.824 0.285 rsl 3047833 DSCAM <0.001 0.906 0.354 rs4871031 DEPDC6 <0.001 0.833 0.225 rs2371438 ERBB4 <0.001 0.893 0.502 rs940539 CDC2L6 <0.001 0.823 0.185 rsl 0120342 PLAA <0.001 0.928 <0.001 rs2836416 ERG 0.001 0.830 0.857 rsl 7826507 PHC3 0.001 0.791 0.080 rsl 7752921 TPM1 0.001 0.850 0.291 rs2186716 ST3GAL4 0.001 0.760 0.574 rs9612266 BCR 0.001 0.897 0.830 rs9559759 COL4A1 0.001 0.923 0.001 rsl 1755592 ZFAND3 0.001 0.807 <0.001 rs3128 CTSH 0.001 0.802 0.023 rs2920836 FRS2 0.001 0.895 0.042 rs4785187 ZNF423 0.001 0.841 0.871 rs7599195 OSBPL6 0.001 0.837 0.463 rs7559527 0SBPL6 0.001 0.919 0.463 rs306410 ATP8A2 0.001 0.931 0.463 rs408359 AGP ATI 0.001 0.887 0.162 rs7386 Cl lorf48 0.001 0.523 0.075 rsl326270 PTPRC 0.001 0.808 0.339 rs765719 ALDH6A1 0.001 0.921 0.005 rs2518523 OR6K6 0.001 0.765 <0.001 rsl 6841047 OR6K6 0.001 0.937 <0.001 rs 1 124922 HIP1 0.001 0.768 0.688 rs2071487 GSTM1 0.001 0.591 <0.001 rs2071487 GSTM1 0.001 0.591 <0.001 rs6415084 LPA 0.001 0.838 0.055 rs4146673 ALK 0.001 0.902 0.835 rs8051232 C0Q7 0.001 0.831 0.871 rsl 1759825 PACSIN1 0.001 0.773 0.285 rsl2991495 DNMT3A 0.001 0.821 0.393 rs6984210 BMP1 0.002 0.819 0.072 rs634370 ABI3 0.002 0.745 0.802 rs2236862 GSTM1 0.002 0.464 <0.001 rs2076109 AP0BEC3F 0.002 0.614 0.064 rsl 1637984 SQRDL 0.002 0.583 0.007 rs2302254 NME1 0.002 0.769 0.234 rsl 0034673 GPRIN3 0.002 0.759 0.657 rsl6962458 NECAB2 0.002 0.786 0.943 rsl 3225749 PTPRZ1 0.002 0.845 0.502 rs4732874 ADRA1 A 0.002 0.879 0.462 rsl 631 1 17 DNAH8 0.002 0.842 0.463 rsl 7062130 GPM6A 0.002 0.719 0.789 rs7098200 ADK 0.002 0.931 0.195 rs7725698 MCTP1 0.002 0.781 0.170 rs3743936 MMP25 0.002 0.703 0.657 rs6141443 RALY 0.002 0.658 0.109 rs7188014 LITAF 0.002 0.778 0.916 rs4558548 PPP1CB 0.002 0.915 0.441 rs3748229 PIK3AP1 0.002 0.770 0.285 rs989128 CACNA1G 0.002 0.785 0.012 rs8129934 AD ARB 1 0.002 0.835 0.495 rs6950693 PTPRN2 0.002 0.572 0.034 rs7557817 FHL2 0.002 0.827 0.422 rsl0486293 HDAC9 0.002 0.941 0.036 rsl 7705427 DNAJC24 0.002 0.913 0.177 rs296886 HNRNPK 0.002 0.881 0.755 rs822570 ARHGEF 1 1 0.002 0.927 0.030 rsl 7294592 SVIL 0.002 0.729 0.574 rs888246 KIAA0999 0.002 0.838 0.463 rs28528975 GAL3ST2 0.002 0.787 0.393 rs2240191 RPH3A 0.002 0.825 0.014 rs2236862 GSTM1 0.002 0.464 O.001 rs2401035 CCDC59 0.002 0.887 0.755 rsl2829066 ITPR2 0.003 0.81 1 0.639 rsl 560489 GPRIN3 0.003 0.902 0.268 rs81 17456 KIF16B 0.003 0.882 0.657 rs7766388 WDR27 0.003 0.849 0.424 rs520328 DSCAML1 0.003 0.780 0.463 rs926561 AKAP12 0.003 0.827 0.128 rsl 0516471 PPP3CA 0.003 0.923 0.846 rsl 3376677 VAV3 0.003 0.875 0.319 rs882422 PCSK6 0.003 0.896 0.657 rs7714610 FSTL4 0.003 0.727 0.143 rs3809449 FAM177A1 0.003 0.660 0.040 rs2523190 GNAI1 0.003 0.804 0.234 rs6556312 RGS14 0.003 0.706 0.586 rsl2761450 AN 3 0.003 0.890 0.778 rs7781413 TMEM195 0.003 0.869 0.943 rs2532007 ADCY9 0.003 0.856 0.424 rs814528 SPTBN4 0.003 0.778 0.079 rsl0929587 ADAM 17 0.003 0.851 0.01 1 rsl0495563 ADAM 17 0.003 0.752 0.011 rs2382553 C9orf93 0.003 0.891 0.285 rs221797 GIGYF1 0.003 0.884 0.203 rs6963037 C7orfl0 0.003 0.805 0.352

Supplementary Table 6 Gene-SNP selection Top 100 list, BMC6-BMC0

Nearest ANOVA p-

SNP Gene value GenTrainScore ChiTestlOO rs7591006 SPAG 16 O.001 0.930 0.655 rsl 51290 KCNQ1 <0.001 0.762 0.463 rs 12624282 C2orf43 <0.001 0.897 0.144 rs2102472 LBH <0.001 0.810 0.349 rs6850131 HSD17B13 <0.001 0.850 0.057 rsl 1735092 HSD17B 13 <0.001 0.842 0.164 rs! 965869 FAM13A <0.001 0.916 0.441 rsl2718455 SNTG2 <0.001 0.837 0.51 1 rs3132680 TRIM31 <0.001 0.919 0.960 rs9827210 CNTN4 <0.001 0.880 0.888 rsl 0451237 R1CH2 <0.001 0.736 0.587 rsl2433712 SRP54 <0.001 0.922 0.891 rs7775864 SNX14 <0.001 0.882 0.425 rs6909767 SNX14 <0.001 0.942 0.425 rs7771612 SNX14 <0.001 0.927 0.425 rs7742691 SNX14 <0.001 0.862 0.425 rs6463016 PRKAP lB <0.001 0.769 0.502 rs l2184386 CUL2 <0.001 0.891 0.267 rsl 7126706 CPNE8 <0.001 0.840 0.234 rs7224186 ARSG <0.001 0.81 1 0.657 rs3129294 HLA-DPB2 <0.001 0.803 0.141 rs l 3072512 FOXP1 <0.001 0.862 0.474 rsl 883414 HLA-DPB2 <0.001 0.929 0.063 rs6577435 CAMTA1 <0.001 NA 0.440 rs6713506 FBXOl l <0.001 0.822 0.871 rs9392366 GMDS <0.001 0.843 0.421 rsl 1219462 VWA5A <0.001 0.897 0.495 rs6454472 SNX14 0.001 0.942 0.549 rs9444352 SNX14 0.001 0.892 0.549 rs2858996 HFE 0.001 0.896 0.441 rs707889 HFE 0.001 NA 0.441 rs989128 CACNA1G 0.001 0.785 0.012 rsl 965869 FAM13A 0.001 0.916 0.441 rs7004524 CSMD1 0.001 0.849 0.018 rs31 18860 DAPK1 0.001 0.844 0.526 rsl2034925 DNAH14 0.001 0.838 0.433 rs7189501 A2BP1 0.001 0.843 0.421 rsl7533945 MY09B 0.001 0.817 0.799 rsl 323080 ClOorfl l 0.001 0.820 0.835 rs6775216 SHOX2 0.001 0.856 0.023 rs31872 PCDHA1 1 0.001 0.822 0.937 rsl3213129 LPAL2 0.001 0.577 0.742 rs9282566 ABCC4 0.001 0.782 0.657 rsl 69250 FLJ25076 0.001 0.792 0.441 rs 17170270 TPK1 0.001 0.913 0.495 rs487269 SRGAP3 0.001 0.762 0.290 rsl7170134 CNTNAP2 0.001 0.935 0.433 rsl 1598750 ADARB2 0.001 0.832 0.006 rs370156 LILRB4 0.001 0.794 0.109 rs6125103 SULF2 0.001 0.837 0.203 rs4671052 EHBP1 0.001 0.842 0.291 rsl 0423215 ZNF347 0.001 0.925 0.290 rs l0814381 RNF38 0.001 0.841 0.354 rsl0824363 ClOorfl l 0.001 0.843 0.015

RP1 - rs6704367 21018.1 0.001 0.900 0.820 rsl7623914 PTPRC 0.001 0.921 0.001 rs6935269 C6orfl 0 0.001 0.867 0.441 rs6594013 ATP2B4 0.001 0.838 0.285 rs2616693 CTN A3 0.001 0.952 0.587 rs2023945 CCDC46 0.001 0.765 0.742 rs3755930 CTBP1 0.002 0.822 0.599 rs7577342 BRE 0.002 0.841 0.672 rs 1902966 BRE 0.002 0.878 0.672 rsl 2465000 BRE 0.002 0.870 0.672 rs6594013 ATP2B4 0.002 0.838 0.285 rsl 062470 CDSN 0.002 0.765 0.667 rsl 867996 CDH23 0.002 0.846 0.039 rsl6955433 CMIP 0.002 0.720 0.040 rsl 1667351 BAX 0.002 0.868 0.234 rsl397202 TAC1 0.002 0.910 0.495 rsl7789420 NPSR1 0.002 0.769 0.943 rsl0149561 F0XN3 0.002 0.830 0.040 rsl 1574 ID3 0.002 0.848 0.462 rs2664371 MMP16 0.002 0.900 0.600 rsl285882 RREB1 0.002 0.787 0.586 rs2071587 F0XN1 0.002 0.730 0.742 rsl 0504965 PGCP 0.002 0.895 0.362 rsl7035482 PEX14 0.002 0.843 0.495 rsl l236172 POLD3 0.002 0.927 0.005 rsl 7443228 IMMP2L 0.002 0.699 0.657 rsl748356 PDSS1 0.002 0.912 0.195 rsl 780179 PDSS1 0.002 0.786 0.195 rsl465314 DTX2 0.002 0.843 0.857 rsl3115520 JAKMIP1 0.002 0.750 0.143 rs7780194 BBS9 0.002 0.792 0.011 rs8009944 RAD51L1 0.002 0.767 0.362 rs4843747 BANP 0.002 0.699 0.495 rs6729843 C2orf43 0.002 0.911 0.185 rs340597 C2orf43 0.002 0.756 0.143 rs2246618 MICB 0.002 0.813 0.891 rs2269058 RNF8 0.002 0.806 0.526 rs4235587 ADCY2 0.002 0.731 0.253 rs9912900 SLC39A1 1 0.002 0.792 0.291 rs 1796236 PTPRN2 0.002 0.685 0.040 rs 1242787 PTPRN2 0.002 0.812 0.320 rs353644 CD44 0.002 0.870 0.267 rs801712 CERK 0.002 0.807 0.586 rsl 7270501 RORA 0.002 0.764 0.657 rs9507557 ATP8A2 0.002 0.919 0.177 rs3471 17 ADAM 10 0.003 0.895 0.548

Supplementary Table 7 Gene-SNP selection Topi 00 list, BMC6-AHC6

Nearest ANOVA p-

SNP Gene value GenTrainScore ChiTestlOO rsl2735646 ARID 1 A <0.001 0.783 0.742 rsl2726287 ARID 1 A <0.001 0.743 0.742 rs 10896623 TIMM10 <0.001 0.924 0.058 rsl 2124339 CAPZB <0.001 0.752 0.352 rs3738919 ITGAV <0.001 0.925 0.595 rsl 51290 KCNQ1 <0.001 0.762 0.463 rs6802942 PPP2R3A <0.001 0.891 0.495 rs4726075 PRKAG2 <0.001 0.717 0.073 rsl 16721 1 1 RDH13 <0.001 0.918 0.424 rsl2034925 DNAH14 <0.001 0.838 0.433 rsl 1699888 SULF2 <0.001 0.862 0.421 rs2857107 HLA-DOB <0.001 0.839 0.143 rs2345122 ZKSCAN2 <0.001 0.824 0.285 rs6775216 SHOX2 <0.001 0.856 0.023 rsl2913832 HERC2 <0.001 0.835 0.023 rs9863749 C3orf20 <0.001 0.854 0.421 rs9913412 ALOX15P <0.001 0.668 0.168 rsl4731 14 NUDCD3 <0.001 0.881 0.235 rsl 800562 HFE O.001 0.787 0.004 rs2252551 C6orfl06 <0.001 0.928 0.399 rs2814998 C6orfl06 <0.001 0.875 0.399 rsl 3379803 AKAP13 <0.001 0.912 0.295 rs676602 NALCN <0.001 0.671 0.857 rsl 3182101 CLTB <0.001 0.900 0.295 rs7959125 ACSS3 <0.001 0.762 0.424 rs9289121 C3orf30 <0.001 0.816 0.960 rsl0889550 LEPR <0.001 0.817 0.502 rsl3213129 LPAL2 0.001 0.577 0.742 rs341397 RORA 0.001 0.765 0.742 RP1 - rs6704367 21018.1 0.001 0.900 0.820 rsl 7003153 FRAS1 0.001 0.878 0.463 rsl 7170270 TPK1 0.001 0.913 0.495 rs6780412 CLDN18 0.001 0.903 0.778 rs467761 1 F0XP1 0.001 0.846 0.137 rs555225 ANK1 0.001 0.714 0.742 rsl6890723 ANK1 0.001 0.704 0.820 rs2518523 OR6K6 0.001 0.765 0.000 rsl6841047 OR6K6 0.001 0.937 0.000 rsl3061519 NLGN1 0.001 0.877 0.657 rs2040784 NFE2L3 0.001 0.691 0.136 rs7521047 NUP210L 0.001 0.919 0.846 rsl 0807151 FKBP5 0.001 0.826 0.058 rs8031 186 ADAMTSL3 0.001 0.874 0.109 rsl 7303530 RORA 0.001 0.920 0.291 rs370156 LILRB4 0.001 0.794 0.109 rs2482424 ABCA1 0.001 0.759 0.290 rsl 0888977 PPAP2B 0.001 0.837 0.399 rs4808571 MY09B 0.001 0.731 0.225 rsl3217929 SYNJ2 0.001 0.842 0.268 rs 17270501 RORA 0.001 0.764 0.657 rs547364 SLC25A24 0.001 0.758 0.799 rs7103581 Cl lorf49 0.001 0.809 0.637 rs7810512 TBRG4 0.001 0.874 0.141 rs2744957 C6orfl06 0.001 0.791 0.094 rs2814992 C6orfl06 0.001 0.938 0.094 rs7235783 SPIRE 1 0.001 0.934 0.506 rsl2516416 AFF4 0.001 0.913 0.422 rsl0988495 COL15A1 0.002 0.801 0.354 rsl71 14699 ANG 0.002 0.820 0.143 rs9726956 FGGY 0.002 0.856 0.029 rs760456 ITGB2 0.002 0.766 0.067 rs2081893 ZNF541 0.002 0.773 0.820 rsl 2972658 ZNF541 0.002 0.800 0.742 rsl2361074 FLJ32810 0.002 0.81 1 0.502 rs6984210 BMP1 0.002 0.819 0.072 rsl 1247287 PCSK6 0.002 0.798 0.260 rsl77181 13 VAT1L 0.002 0.809 0.688 rsl418253 LPHN2 0.002 0.868 0.424 rs367881 LPHN2 0.002 0.874 0.421 rs4491236 NTM 0.002 0.854 0.899 fsl 7133676 OGDH 0.002 0.849 0.820 rs2023945 CCDC46 0.002 0.765 0.742 rsl531817 PCSK6 0.002 0.785 0.141 rsl 573994 ITPR2 0.002 0.845 0.285 rs3790515 RORC 0.002 0.775 0.495 rs3859534 LILRA6 0.002 0.737 0.441 rsl 1259333 FAM107B 0.002 0.693 0.144 rsl 1967633 TMEM63B 0.002 0.710 0.014 rs2071587 FOXN1 0.002 0.730 0.742 rsl 3225749 PTPRZ1 0.002 0.845 0.502 rs306410 ATP8A2 0.002 0.931 0.463 rs4775310 RORA 0.002 0.736 0.014 rs320109 RCOR2 0.002 0.81 1 0.295 rsl 55104 ITGA4 0.003 0.921 0.891 rsl 1208660 LEPR 0.003 0.895 0.295 rs625014 RAB31 0.003 0.818 0.164 rsl0071707 PDZD2 0.003 0.919 0.587 rs222857 CLDN7 0.003 0.807 0.234 rsl2582168 NCOR2 0.003 0.817 0.051 rsl 12544 LZTR1 0.003 0.870 0.128 rsl415701 L3MBTL3 0.003 0.862 0.724 rs995435 TGFBR2 0.003 0.843 0.393 rs7770046 TMEM181 0.003 0.781 0.185 rs3751909 FOXK2 0.003 0.731 0.352 rsl 7128050 GCH1 0.003 0.899 0.463 rsl 0423215 ZNF347 0.003 0.925 0.290 rs2186716 ST3GAL4 0.003 0.760 0.574

RP1 1 - rsl 0989419 35N6.1 0.003 0.859 0.058 rs7781464 CNTNAP2 0.003 0.876 0.135 rs2238202 RGS6 0.003 0.925 0.295

Supplementary Table 8 Gene-SNP selection Topi 00 list, (BMC6-BMC0)-(AHC6-AHC0)

Nearest ANOVA p-

SNP Gene value GenTrainScore ChiTestlOO

rs7210402 SGSM2 <0.001 0.839 0.863

rsl 806516 P2RY6 <0.001 0.694 0.914

rs2345122 ZKSCAN2 <0.001 0.824 0.285

rs7004524 CSMD1 <0.001 0.849 0.018

rsl7817463 DISCI <0.001 0.733 0.421

rs 1 1623922 KCNK13 <0.001 0.751 0.064

rs370133 NRCAM <0.001 0.880 0.51 1

rs341397 RORA O.001 0.765 0.742

rs2427638 PCMTD2 <0.001 0.824 0.064

rs7224186 ARSG <0.001 0.81 1 0.657

rsl 16721 1 1 RDH13 <0.001 0.918 0.424

rs 10889550 LEPR <0.001 0.817 0.502

rs7203078 CMIP <0.001 0.753 0.574

rsl2451892 SGSM2 <0.001 0.826 0.141

rs7203568 WWOX <0.001 0.815 0.778

rs451 1641 RTN2 0.001 0.799 0.339

rsl 3379803 AKAP13 0.001 0.912 0.295

rsl0888977 PPAP2B 0.001 0.837 0.399

rs2343869 SSPN 0.001 0.845 0.318

rs845204 CAMTA1 0.001 0.930 0.063

rsl 1079323 MSI2 0.001 0.928 0.009

rs3738919 ITGAV 0.001 0.925 0.595

rsl 2460755 1NSR 0.001 0.927 0.433

rs7235783 SPIRE 1 0.001 0.934 0.506

rsl 0071707 PDZD2 0.001 0.919 0.587

rsl 018788 LARGE 0.001 0.895 0.474

rs4335165 MTUS1 0.001 0.753 0.287

rs389883 STK19 0.001 0.902 0.522

rs4801 163 ZNF667 0.001 0.936 0.177 rs9304776 ZNF667 0.001 0.852 0.177 rsl 3225749 PTPRZ1 0.001 0.845 0.502 rs4384073 DDX58 0.001 0.778 0.203 rs2836416 ERG 0.001 0.830 0.857 rs9346818 LPAL2 0.001 0.891 0.522 rs3804267 PPAP2A 0.001 0.918 0.830 rsl2246732 FAM107B 0.001 0.869 0.225 rs6785790 SETD2 0.001 0.854 0.048 rsl 6869706 SLIT2 0.001 0.897 0.655

RP1 1- rsl0989419 35N6.1 0.001 0.859 0.058 rs9853081 F0XP1 0.001 0.914 0.001 rs7712431 CSNK1A1 0.001 0.932 0.655 rs6415084 LPA 0.002 0.838 0.055 rs7076232 BTBD16 0.002 0.805 0.587 rsl l 814901 BTBD16 0.002 0.835 0.587 rs2481665 INADL 0.002 0.815 0.003 rs7625067 SETD2 0.002 0.813 0.009 rs2071587 FOXN1 0.002 0.730 0.742 rsl 0426628 SULT2B 1 0.002 0.677 0.871 rs3893677 KCTD1 0.002 0.809 0.891 rs2010010 GALNT10 0.002 0.777 0.090 rs2176771 MMP16 0.002 0.675 0.080 rsl 2034925 DNAH14 0.002 0.838 0.433 rsl7170270 TPK1 0.002 0.913 0.495 rs9390569 SASH1 0.002 0.930 0.506 rsl 1208660 LEPR 0.002 0.895 0.295 rs l64577 SLC30A5 0.002 0.840 0.001 rsl69250 FLJ25076 0.002 0.792 0.441 rs2260000 BAT2 0.002 0.810 0.526 rs2736172 BAT2 0.002 0.728 0.526 rsl0814381 RNF38 0.002 0.841 0.354 rsl l33195 MXI1 0.002 0.905 0.291 rs2298229 0LFM4 0.002 0.909 0.399 rs 10979586 IKBKAP 0.002 0.935 0.260 rsl883414 HLA-DPB2 0.002 0.929 0.063 rs2371438 ERBB4 0.002 0.893 0.502 rs2010576 MICAL2 0.002 0.819 0.549 rs550338 S0X5 0.002 0.879 0.043 rs788332 YH14 0.002 0.789 0.138 rs9726956 FGGY 0.002 0.856 0.029 rs8087174 0SBPL1A 0.002 0.869 0.639 rsl51290 KCNQ1 0.002 0.762 0.463 rs3094476 KCTD5 0.002 0.850 0.001 rs876687 TGFBR2 0.002 0.847 0.502 rs3773661 TGFBR2 0.002 0.794 0.495 rs6775216 SHOX2 0.003 0.856 0.023 rs7901290 CAMK1D 0.003 0.900 0.672 rs3809572 SMAD3 0.003 0.726 0.495 rs2186716 ST3GAL4 0.003 0.760 0.574 rsl 1967633 TMEM63B 0.003 0.710 0.014 rs6925303 FY 0.003 0.882 0.019 rs6914091 FYN 0.003 0.713 0.019 rs6930230 FYN 0.003 0.933 0.019 rs555225 ANK1 0.003 0.714 0.742 rsl 6890723 ANK1 0.003 0.704 0.820 rsl 185331 1 SLC03A1 0.003 0.799 0.440 rs6650615 MPPE1 0.003 0.916 0.290 rsl l33195 MX11 0.003 0.905 0.291 rs2286294 GLI3 0.003 0.959 0.137 rsl 7799872 ADCY3 0.003 0.746 0.421 rs2744805 RIMS3 0.003 0.785 0.614 rs3016562 PARK2 0.003 0.852 0.009 rs6868292 PPAP2A 0.003 0.876 0.290 rs 16924332 ABCC9 0.003 0.937 0.778 rs2201945 PCDH7 0.003 0.844 0.295 rsl0010739 PCDH7 0.003 0.899 0.030 rs2285431 HDAC9 0.003 0.837 0.360 rsl0503284 CSMD1 0.003 0.719 0.143 rs3774491 CACNA1D 0.003 0.839 0.888 rs2518523 OR6K6 0.003 0.765 <0.001 rsl 6841047 OR6K6 0.003 0.937 <0.001