Integrative Analysis Of Multi-Omics Data For Obesity Genetic Risk Factors In America Caucasian Female

BackgroundObesity is a complex,multifactorial condition in which genetic factors play an important role.Most of the systematic studies currently are focused on single or two omics measurements such as DNA,RNA or metabolite level individually,few knowledge could be obtained about the cross-talks between various omics levels and the underlying biological networks that drive complex phenotypes.Few studies focused on the female populations.With the advance of emerging high-throughput sequencing technology such as whole genome sequencing?WGS?,RNA-sequencing?RNA-Seq?,reduced-representation bisulfite sequencing?RRBS?,and liquid chromatography-mass spectrometry?LC-MS?,multi-omics data including genomics,epigenomics,transcriptomics and metabolomics are rapidly generated.As a result,more and more researchers are currently working on the integration of comprehensive multi-omics data to create new and meaningful biological knowledge,but those studies focused on obesity are rare.Peripheral blood monocytes?PBMs?have been increasingly used as surrogate tissue for transcriptomic and epigenomic biomarkers researches because of their modulation mechanisms in relation to obesity risk.Therefore,we will use PBMs as an example cell type for illustration for the first integrative multi-omics study of obesity in Caucasian females.Objective1.To identify DEGs,DMRs and DAMs for different omics level between group overweight/obesity and normal weight.2.To identify the correlation status among multi-omics data and to identify expression QTLs?eQTLs?,methylation QTLs?meQTLs?and metabolites QTLs?metaQTLs?for each omics individually.3.To identify the potential causal pairs and mediation effect among different omics.Methods1.According to WHO definition for obesity,we recruited Caucasian females using overweight/obesity and normal weight 1:1 design.Whole genome sequencing?WGS?,reduced-representation bisulfite sequencing?RRBS??for peripheral blood monocytes,PBM?,RNA-sequencing?RNA-seq??PBM?,and liquid chromatography-mass spectrometry?LC-MS?were performed for the DNA,RNA and serum bio-samples respectively to generate the multi-omics data including genomics,epigenomics,transcriptomics and metabolomics.2.For RNA-seq data,we performed DE analysis identify the DE genes?DEGs?between two groups,those significant DEGs were subject to the Multiscale Embedded Gene Co-expression Network Analysis?MEGENA?to identify functional co-expressed gene modules and hub genes associated with obesity;For RRBS data,we adopted Logistic regression to identify the differentially methylated regions/bases?DMRs?between two groups;For the LC-MS data,we conducted both partial least squares regression-discriminant analysis?PLS-DA?and logistic regression analysis to detect the differentially accumulated metabolites?DAMs?between two groups.3.We first performed Spearman correlation analysis among hub genes,DMRs and DAMs.Then,by performing QTL analysis,eQTL,meQTL and metaQTL datasets were generated for further MR analysis.Next,we applied the bi-directional MR approach to hub genes and DMRs,DMRs and DAMs,hub genes and DAMs to detect the potential casual pairs among multi-omics data.4.To further integrate those multi-omics data for BMI and decipher the crosstalk among them,we applied network MR analysis to identify the potential causal pairs which have the mediation effect.Results1.There were 104 Caucasian females were included in the current study:overweight/obesity 52 vs normal weight 52.2.For RNA-Seq data,we identified 214 DE genes?adjP<0.01?,then those genes were classified into 17 gene modules and 8 genes were prioritized as the hub genes,where six of them were previously reported,two were novel?LUZP6 and PLCB2?.For RRBS data,by using the threshold of Q-value<0.01 and differential methylation percentage?DMP?larger than 10%,we totally identified 95 DMRs,then those 95 DMRs were annotated to 67 nearest genes.When using more stringent threshold of Q<0.01 and DMP>15%,14 DMRs were prioritized,for the 12 genes those DMRs annotated to,four of them were previously reported,the rest six genes were novel?PACRG,LINC00494,KLHL4,DTX1,VCX3A and VSTM1?.For LC-MS data,we identified 12 DAMs for obesity,ten of them were previously reported,two were novel?Indole-3-acetate and N-methyl-D-aspartic acid?.3.Spearman correlation analysis demonstrates significant correlation among different omics.QTL analysis identified 3560 eQTL,734 meQTL and 9055 metaQTL for hub DEGs,DMRs and DAMs,individually.By using MR analysis,we successfully detected 7 significant causal pairs between hub genes and DMRs,40 significant causal pairs between DMRs and DAMs and 42 significant causal pairs between hub genes and DAMs.4.Network MR identified 18 potential causal pairs which have mediation effect?Isobutyrylcarnitine_ANO6₆.110721178,Plasmenyl-LysoPE_ANO6₆.110721178,3-?2-Hydroxyphenyl?Propanoate_ANO6₆.110721178 etc.?,and 20 biomarkers were included,17 of them were previously reported,the rest three were novel?PA CRG-AS1 annotated by DMR 6.163743051,Indole-3-acetate and N-methyl-D-aspartic acid?.Conclusions1.Differentially analysis has the potentially to identify the unique significant biomarkers for each omics individually.2.Multi-omics integration has the potential to capture complementary effect and the interactions between different omics levels.In the current study,through the application of QTL,MR and network MR,we were able to identify the correlation,potential causal association and the crosstalk among different omics.Finally,we detected 18 potential causal biomarkers pairs which have mediation effect,there were 20 biomarkers included,17 of them were previously reported,while the other three were novel,which provide target biomarker for future obesity related biological experiments and provide solid foundation for future precision medicine and targeted treatment.