ASHG_2015

Talk I gave in "Computing Functional Variants" session at the annual meeting of the American Society for Human Genetics (ASHG) in Baltimore on 10/10/2015.

Transcript

1. Integrating genome-wide association and co-expression network data for novel BMD gene discovery Charles R. Farber, Larry Mesner, Joseph P. Stains, Steven M. Tommasini, Mark C. Horowitz, Clifford J. Rosen, Gina Calabrese American Society of Human Genetics Annual Meeting October 10th, 2015
2. • Genome-Wide Association Studies (GWAS) have identified thousands of associations 1. Gene(s) 2. Function
3. 1. Genes that influence a disease are often functionally related (Goh, K.-I. et al. 2007. PNAS) 2. Functionally related genes are often co-expressed (Nayak, R. R. et al. 2009. Genome Research) Observations
4. Gene X Gene Y Gene Z Gene A Gene B Gene C Non-Causal Causal GWAS loci
5. Gene X Gene Y Gene Z Gene A Gene B Gene C Non-Causal Causal GWAS loci Functionally related and co- expressed
6. Gene X Gene Y Gene Z Gene A Gene B Gene C Non-Causal Causal GWAS loci Functionally related and co- expressed Genes located within GWAS loci
7. Gene X Gene Y Gene Z Gene A Gene B Gene C Non-Causal Causal GWAS loci Functionally related and co- expressed Genes located within GWAS loci
8. Enriched co-expression modules Gene X Gene Y Gene Z Gene A Gene B Gene C Non-Causal Causal GWAS loci Functionally related and co- expressed Genes located within GWAS loci
9. Enriched co-expression modules Gene X Gene Y Gene Z Gene A Gene B Gene C Non-Causal Causal GWAS loci Functionally related and co- expressed Genes located within GWAS loci Gene X Gene Y Gene Z Prioritize Genes Predict Function
10. 64 SNPs P<5.0 x 10-8 234 Genes Bone transcriptomic profiles from male mice of 96 inbred mouse strains (Calabrese et al. 2012. PLoS Genetics) Genome-wide co-expression network GEFOSII BMD GWAS (N~80K) (Estrada, K., et al. 2012. Nature Genetics)
11. GWAS gene enrichment in modules 6 and 9 0 1 2 3 4 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 Module −log10(Enrichment P-value) 6 9
12. GWAS gene enrichment in modules 6 and 9 0 1 2 3 4 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 Module −log10(Enrichment P-value) 6 9 • Enriched for genes with well- known roles in osteoblasts • Preferentially expressed in osteoblast-lineage cells • Module 6 and 9 genes, as a group, were co- expressed
13. Osteoblast Functional Module (OFM) GWAS SNP Chr Mbp Gene rs1053051 12 107.4 TMEM263 rs4233949 2 54.7 SPTBN1 rs4790881 17 2.1 SMG6 rs3736228 11 68.2 PPP6R3 rs4790881 17 2.1 HIC1 rs13245690 7 120.8 CPED1 rs7071206 10 79.4 DLG5 rs7851693 9 133.5 FUBP3 rs3801387 7 121.0 FAM3C rs11623869 14 103.9 MARK3 rs7071206 10 79.4 KCNMA1 GWAS SNP Chr Mbp Gene rs11755164 6 44.6 RUNX2 rs2016266 12 53.7 SP7 rs2062377 8 120.0 TNFRSF11B rs7932354 11 46.7 LRP4 rs3736228 11 68.2 LRP5 rs4792909 17 41.8 SOST rs1373004 10 54.4 DKK1 rs2887571 12 1.6 WNT5B rs7521902 1 22.5 WNT4 rs13245690 7 120.8 WNT16 rs12407028 1 68.6 WLS rs227584 17 42.2 HDAC5 GWAS SNP Chr Mbp Gene rs6532023 4 88.8 MEPE rs4792909 17 41.8 MEOX1 rs1366594 5 88.4 MEF2C rs7932354 11 46.7 MDK rs1566045 16 51.0 CYLD rs7017914 8 71.6 EYA1 rs479336 1 172.2 DNM3OS rs1346004 2 166.6 GALNT3 rs7108738 11 15.7 INSC rs736825 12 54.4 HOXC8 rs736825 12 54.4 HOXC6 1. Module 6 and 9 genes 2. Genes whose expression was correlated (P<0.001) with the eigengene of either module
14. OFM SNPs overlap active regulatory elements
15. SNP Group OFM Lead SNPs (N=30) NOFM Lead SNPs (N=34) OFM matched random SNP set OFM SNPs overlap active regulatory elements
16. SNP Group OFM Lead SNPs (N=30) NOFM Lead SNPs (N=34) OFM matched random SNP set OFM SNPs overlap active regulatory elements 0 10 20 30 40 50 H3K27ac Overlap (%) P=0.02 P<0.001 H3K4me2 P=0.02 P<0.001 Primary Human Osteoblasts ENCODE Project Consortium. (2014). Nature.
17. OFM histone overlap is specific to osteoblasts ADRENAL BLOOD BONE BRAIN BREAST CERVIX ESC ESC_DERIVED FAT GI_COLON GI_DUODENUM GI_ESOPHAGUS GI_INTESTINE GI_RECTUM GI_STOMACH HEART IPSC LIVER LUNG MUSCLE MUSCLE_LEG OVARY PANCREAS PLACENTA SKIN SPLEEN STROMAL_CONNECTIVE THYMUS VASCULAR OSTEOBLASTS CHONDROCYTES MESENCHYMAL STEM CELLS FIBROBLASTS 10 20 30 40 −0.1 0.0 0.1 0.2 0.3 Difference (% OFM overlap - % NOFM overlap) Overlap (%) H3K27ac Roadmap Epigenomics Consortium. (2015). Nature.
18. Osteoblast Functional Module (OFM) GWAS SNP Chr Mbp Gene rs1053051 12 107.4 TMEM263 rs4233949 2 54.7 SPTBN1 rs4790881 17 2.1 SMG6 rs3736228 11 68.2 PPP6R3 rs4790881 17 2.1 HIC1 rs13245690 7 120.8 CPED1 rs7071206 10 79.4 DLG5 rs7851693 9 133.5 FUBP3 rs3801387 7 121.0 FAM3C rs11623869 14 103.9 MARK3 rs7071206 10 79.4 KCNMA1 GWAS SNP Chr Mbp Gene rs11755164 6 44.6 RUNX2 rs2016266 12 53.7 SP7 rs2062377 8 120.0 TNFRSF11B rs7932354 11 46.7 LRP4 rs3736228 11 68.2 LRP5 rs4792909 17 41.8 SOST rs1373004 10 54.4 DKK1 rs2887571 12 1.6 WNT5B rs7521902 1 22.5 WNT4 rs13245690 7 120.8 WNT16 rs12407028 1 68.6 WLS rs227584 17 42.2 HDAC5 GWAS SNP Chr Mbp Gene rs6532023 4 88.8 MEPE rs4792909 17 41.8 MEOX1 rs1366594 5 88.4 MEF2C rs7932354 11 46.7 MDK rs1566045 16 51.0 CYLD rs7017914 8 71.6 EYA1 rs479336 1 172.2 DNM3OS rs1346004 2 166.6 GALNT3 rs7108738 11 15.7 INSC rs736825 12 54.4 HOXC8 rs736825 12 54.4 HOXC6
19. Chr14q32.32 GWAS locus Estrada, K. et al. (2012). Nature Genetics. 0 1 2 3 4 5 6 7 rs11623869 • • •• • • • • • • • • • • • • • • • • •• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •• • • • • • • • • • • •• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •• • • • • • • • • •• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •• • • • • • • • • • •• • • • • • • • • • • • • • 0.2 0.4 0.6 0.8 r2 GEFOSII Femoral Neck BMD EIF5 SNORA28 MARK3 CKB TRMT61A BAG5 APOPT1 103.75 103.8 103.85 103.9 103.95 104 Position on chr14 (Mb) -log10(P-value)
20. Chr14q32.32 GWAS locus Estrada, K. et al. (2012). Nature Genetics. 0 1 2 3 4 5 6 7 rs11623869 • • •• • • • • • • • • • • • • • • • • •• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •• • • • • • • • • • • •• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •• • • • • • • • • •• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •• • • • • • • • • • •• • • • • • • • • • • • • • 0.2 0.4 0.6 0.8 r2 GEFOSII Femoral Neck BMD EIF5 SNORA28 MARK3 CKB TRMT61A BAG5 APOPT1 103.75 103.8 103.85 103.9 103.95 104 Position on chr14 (Mb) -log10(P-value) 9.00 9.25 9.50 9.75 10.00 0.06 0.07 0.08 0.09 0.10 Femoral BMD (g/cm2) Bone Mark3 Expression (log 2 (INT)) r=-0.25, P=0.036
21. Chr14q32.32 GWAS locus Estrada, K. et al. (2012). Nature Genetics. 0 1 2 3 4 5 6 7 rs11623869 • • •• • • • • • • • • • • • • • • • • •• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •• • • • • • • • • • • •• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •• • • • • • • • • •• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •• • • • • • • • • • •• • • • • • • • • • • • • • 0.2 0.4 0.6 0.8 r2 GEFOSII Femoral Neck BMD EIF5 SNORA28 MARK3 CKB TRMT61A BAG5 APOPT1 103.75 103.8 103.85 103.9 103.95 104 Position on chr14 (Mb) -log10(P-value) 9.00 9.25 9.50 9.75 10.00 0.06 0.07 0.08 0.09 0.10 Femoral BMD (g/cm2) Bone Mark3 Expression (log 2 (INT)) r=-0.25, P=0.036 Mark3 Osteoblast activity & BMD
22. Primary calvarial osteoblasts Mark3 siRNAs Differentiate Quantify Mineralization SC Mark3_1 Mark3_2 0.09 0.12 0.15 0.18 0.21 Alizarin Red (mM) SC Mark3 1 Mark3 2 P<0.05 0.00 0.25 0.50 0.75 1.00 SC Mark3 1 Mark3 2 Mark3 Relative Expression P<0.01
23. Increased Femoral BMD in Mark3 Gene Trap
24. Increased Femoral BMD in Mark3 Gene Trap KOMP Gene Trap (tm1a) targeted ES Cells +/tm1a 45% reduction in Mark3 expression
25. Increased Femoral BMD in Mark3 Gene Trap Femoral BMD 12 week Males Genotype Adjusted BMD (g/cm2) 0.066 0.068 0.070 0.072 0.074 0.076 +/+ +/tm1a N=13 N=17 P=0.01 KOMP Gene Trap (tm1a) targeted ES Cells +/tm1a 45% reduction in Mark3 expression
26. http://www.gtexportal.org/; GTEx Consortium. (2015). Science. 0 5 10 15 rs11623869 • • • • • • • • • • • • • • • • •• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •• • • • • • • • • • • • • • • 0.2 0.4 0.6 0.8 r2 EIF5 SNORA28 MARK3 CKB TRMT61A BAG5 APOPT1 103.75 103.8 103.85 103.9 103.95 104 Position on chr14 (Mb) GTEX MARK3 THYROID eQTL -log10(P-value) MARK3 local eQTL
27. http://www.gtexportal.org/; GTEx Consortium. (2015). Science. 0 5 10 15 rs11623869 • • • • • • • • • • • • • • • • •• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •• • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • • •• • • • • • • • • • • • • • • 0.2 0.4 0.6 0.8 r2 EIF5 SNORA28 MARK3 CKB TRMT61A BAG5 APOPT1 103.75 103.8 103.85 103.9 103.95 104 Position on chr14 (Mb) GTEX MARK3 THYROID eQTL Prostate Thyroid Adrenal Gland Artery Coronary Colon Transverse Pancreas Skin Sun Exposed Stomach Adipose Subcutaneous Muscle Skeletal Lung BMD 0.00 0.25 0.50 0.75 Effect of the rs11623869 “T” allele 4.6 x 10-2 5.2 x 10-16 7.4 x 10-3 4.2 x 10-2 1.1 x 10-2 9.9 x 10-3 3.6 x 10-2 2.9 x 10-2 1.7 x 10-2 1.4 x 10-2 1.6 x 10-13 1.1 x 10-2 -log10(P-value) MARK3 local eQTL
28. Conclusions • We identified the OFM by overlaying GWAS information on a disease-relevant co-expression network • We predict that the OFM is enriched for genes responsible for BMD GWAS loci and can be used to guide future experimental work • MARK3 local eQTL likely underlies the BMD association on Chr14q32.32
29. Acknowledgements: Funding CRF: NIH/NIAMS R01-AR057759 NIH/NIAMS R56-AR064790 NIH/NIAMS R01-AR064790 University of Virginia, Center for Public Health Genomics CJR:NIH/NIDDK R24-DK092759 NIH/NIGMS P30-GM106391 NIH/NIGMS P30-GM103392