R/~old/README.md

In BaderLab/POPPATHR: Population-based pathway analysis of SNP-SNP coevolution

PopulationPathways

The same pathways are present in all humans, but are these genes somehow tuned differently based on population ancestry?

This is a project to develop a computational pipeline that determines instances of population-driven SNP-SNP coevolution at the pathway level in an effort to better understand the evolution of human pathways. A wide number of software packages and statistical methods are used and are outlined below (NOTE: the following pipeline is currently carried out between the CEU and YRI population cohorts)

Methods summary

| METHOD | RESULT | | ------------- | ------------- | | PLINK | ~1.5 million SNPs genotyped in CEU and YRI (HapMap 3) | | GSEA | 56 pathways (19 non-redundant) enriched for population-driven positive selection (i.e., high FST genes) | | Linkage disequilibrium | 4 (CEU) / 10 (YRI) within-pathway and 16 (CEU) / 59 (YRI) between-pathway coevolution signals discoered (FDR ≤ 0.2)|

HYPOTHESIS: evolutionary maintenance of pathway-level SNP interactions that influence population fitness

Steps of computational pipeline:

Pipeline scripts found here

# within R environment (calls recodeFAM.R, popPCA.R, calcFST.R, SNP2gene.R, setupGSEArun.R, getPathStats.R, LDstatsWPM.R, and LDstatsBPM.R)
> source(runPipeline.R) # runs entire pipeline

1) PLINK

• PLINK 1.90 for OS X downloaded here
• Pipeline compatible with SNP genotyping data
• 1,423,541 polymorphic SNPs from the International HapMap 3 Project
  • Converted from MAP/PED to BIM/BED/FAM format
• Genotypes required for two population comparative analyses
  • Original: CEU vs. YRI
  • Replication: CEU vs. LWK
• Raw PLINK data on Catherine's laptop ~/PopulationPathways/data/HM3_2010-05_phase3
• 3 purposes:
• Recode PLINK fam file to case/control format by population (recodeFAM.R)
• Perform PCA on tested populations (popPCA.R)
• Calculate SNP-level FST as a marker for population-driven positive selection (calcFST.R)

2) GSEA (Gene-set enrichment analysis)

• Perl scripts from GenGen software tool package

Input:

• Pathway annotation file (downloaded from Bader lab website)
• File used here found in git repo: Human_GOBP_AllPathways_no_GO_iea_April_24_2016_symbol.gmt
• SNP-level FST statistic file (calcFST.R)
• NOTE: the second column labelled 'CHI2' for purposes of running GenGen GSEA script as it originally expects GWAS p values or CHI2, represents FST) ```

  head(markerFST) Marker CHI2 1 rs1000000 0.100100 2 rs10000003 0.122500 3 rs10000007 -0.273025 4 rs10000010 -0.005300 5 rs10000011 -0.179690 6 rs10000015 -0.177080 ```

• SNP-to-gene mapping file (SNP2gene.R)
• SNPs are mapped to their physically nearest gene
• GSEA is set to filter out any mapping with SNPs lying >500kb from its nearest gene ```

  head(snp2gene) V1 V2 V3 1 rs12562034 LOC100288069 0 2 rs7518545 LOC100288069 0 3 rs9651273 AGRN 0 4 rs61766341 AGRN 0 5 rs12134754 AGRN 0 6 rs11586034 AGRN 0 ```

Output:

• Table containing pathway enrichment results

> head(res)
Geneset
1                                                     CELL FATE COMMITMENT%GOBP%GO:0045165
2              SRP-DEPENDENT COTRANSLATIONAL PROTEIN TARGETING TO MEMBRANE%GOBP%GO:0006614
3                 EUKARYOTIC TRANSLATION TERMINATION%REACTOME DATABASE ID RELEASE 56%72764
4                 EUKARYOTIC TRANSLATION ELONGATION%REACTOME DATABASE ID RELEASE 56%156842
5                                           VIRAL MRNA TRANSLATION%REACTOME%R-HSA-192823.1
6 TRANSMEMBRANE RECEPTOR PROTEIN SERINE/THREONINE KINASE SIGNALING PATHWAY%GOBP%GO:0007178
Size    ES   NES NominalP   FDR   FWER
1  109 0.406 4.667   0.0000 0.009 0.0506
2   83 0.130 4.629   0.0007 0.009 0.0556
3   77 0.114 4.592   0.0025 0.009 0.0623
4   78 0.107 5.024   0.0019 0.010 0.0184
5   75 0.126 4.901   0.0013 0.010 0.0260
6  134 0.372 4.676   0.0000 0.010 0.0492

Results:

• CEU-YRI: 84 pathways significantly enriched for high-FST genes (FDR ≤ 0.05)
• CEU-LWK: 130 pathways significantly enriched for high-FST genes (FDR ≤ 0.05)
• Replication by both analyses reveals 56 'selection-enriched' pathways show true evidence of population-driven positive selection
• In other words, SNPs show significant differentiation in FST between European and African populations at the pathway level
• Immune, metabolism, and development themes
• Generated binary PLINK files for downstream analysis (getPathStats.R)
• Selection-enriched pathways: 56 (19 non-redundant)
• Unenriched pathways: 53 (i.e., SNP-level FST values are not significantly differentiated between Europeans and Africans, negative controls)

3) Linkage disequilibrium (LD)

• Reference snpStats LD vignette
• Inter-chromosomal LD (r2) used as a proxy to detect SNP-SNP coevolution at the pathway level
• Two pathway-level interaction models implemented (motivated by recent BridGE method):
• Within-pathway -- detecting SNP-SNP coevolution within a single pathway
• Between-pathway -- detecting SNP-SNP coevolution between pairs of pathways (functional pathway clusters)

Input:

• SNP genotype and location data per selection enriched and unenriched pathway (getPathStats.R)

Output:

• LD r2 value per inter-chromosomal pairwise SNP combination specific to population cohort
• Within-pathway results ```

  head(dat) snp_1 chr_1 pos_1 snp_2 chr_2 pos_2 R.squared pathway 1 rs10119465 9 19376144 rs1693547 8 101714394 8.019353e-03 pathway_1 2 rs10119465 9 19376144 rs1881157 4 99712077 5.271225e-03 pathway_1 3 rs10119465 9 19376144 rs11684944 2 3633917 1.151087e-02 pathway_1 4 rs10119465 9 19376144 rs632401 7 73563425 3.049999e-03 pathway_1 5 rs10119465 9 19376144 rs6545480 2 55460751 2.641132e-02 pathway_1 6 rs10119465 9 19376144 rs11586570 1 93305371 1.475042e-05 pathway_1 set pop 1 Enriched CEU 2 Enriched CEU 3 Enriched CEU 4 Enriched CEU 5 Enriched CEU 6 Enriched CEU ```

• Between-pathway results ```

  head(dat) snp_1 chr_1 pos_1 gene_1 snp_2 chr_2 pos_2 R.squared 1 rs166125 5 65201373 ERBB2IP rs10762276 10 71052469 0.0084093756 2 rs12493155 3 138400332 PIK3CB rs10762276 10 71052469 0.0079737336 3 rs2131432 17 1297472 YWHAE rs10762276 10 71052469 0.0102617490 4 rs12076073 1 154944156 SHC1 rs10762276 10 71052469 0.0002102607 5 rs741334 12 112985328 PTPN11 rs10762276 10 71052469 0.0031810952 6 rs3756668 5 67596088 PIK3R1 rs10762276 10 71052469 0.0048012803 pathway_pair1 1 ALPHA6BETA4INTEGRIN_IOB_ALPHA6BETA4INTEGRIN 2 ALPHA6BETA4INTEGRIN_IOB_ALPHA6BETA4INTEGRIN 3 ALPHA6BETA4INTEGRIN_IOB_ALPHA6BETA4INTEGRIN 4 ALPHA6BETA4INTEGRIN_IOB_ALPHA6BETA4INTEGRIN 5 ALPHA6BETA4INTEGRIN_IOB_ALPHA6BETA4INTEGRIN 6 ALPHA6BETA4INTEGRIN_IOB_ALPHA6BETA4INTEGRIN pathway_pair2 ixn_num set pop 1 CARBOHYDRATE_TRANSPORT_GOBP_GO_0008643 interaction_1 Enriched CEU 2 CARBOHYDRATE_TRANSPORT_GOBP_GO_0008643 interaction_1 Enriched CEU 3 CARBOHYDRATE_TRANSPORT_GOBP_GO_0008643 interaction_1 Enriched CEU 4 CARBOHYDRATE_TRANSPORT_GOBP_GO_0008643 interaction_1 Enriched CEU 5 CARBOHYDRATE_TRANSPORT_GOBP_GO_0008643 interaction_1 Enriched CEU 6 CARBOHYDRATE_TRANSPORT_GOBP_GO_0008643 interaction_1 Enriched CEU ```

• Pathway-level FDR statistic representing the significance of coevolution signal (determined via Kolmogorov-Smirnov statistic)

• Within-pathway: LD r2 distribution per selection-enriched pathway against cumulative LD r2 distribution of unenriched pathways head(pvals) pathway 1 EMBRYONIC_MORPHOGENESIS 2 CHROMATIN_SILENCING 3 NONSENSE_MEDIATED_DECAY_NMD_INDEPENDENT_OF_THE_EXON_JUNCTION_COMPLEX_EJC 4 NONSENSE_MEDIATED_DECAY_NMD_ENHANCED_BY_THE_EXON_JUNCTION_COMPLEX_EJC 5 NONSENSE_MEDIATED_DECAY_NMD 6 CELL_CHEMOTAXIS pvals_CEU bonf_CEU fdr_CEU pvals_YRI bonf_YRI fdr_YRI 1 0.00142 0.0793 0.0793 0.26100 1.000 0.3570 2 0.00367 0.2058 0.1029 0.80100 1.000 0.8790 3 0.01135 0.6357 0.1325 0.02480 1.000 0.0555 4 0.01183 0.6623 0.1325 0.00289 0.162 0.0192 5 0.01183 0.6623 0.1325 0.00289 0.162 0.0192 6 0.02210 1.0000 0.1920 0.00253 0.141 0.0192

  • NOTE: functionally redundant pathways are reported as a single signal of within-pathway coevolution (based on thematic functional grouped as determined by Cytoscape EnrichmentMap)

• Between-pathway: LD r2 distribution per selection-enriched pathway-pathway pair against cumulative LD r2 distribution of all unenriched pathway-pathway pairs head(pvals) interaction.pathway_pair1 1 ALPHA6BETA4INTEGRIN 2 EMBRYONIC_MORPHOGENESIS 3 ALPHA6BETA4INTEGRIN 4 ALPHA6BETA4INTEGRIN 5 CELLULAR_RESPONSE_TO_LIGHT 6 ALPHA6BETA4INTEGRIN interaction.pathway_pair2 1 CELLULAR_RESPONSE_TO_LIGHT 2 IL4_MEDIATED_SIGNALING_EVENTS 3 PROTEASOMAL_PROTEIN_CATABOLIC_PROCESS 4 VIRAL_TRANSLATION_AND_PROTEIN_TARGETING 5 IMMUNE_CELL_REGULATION 6 CELL_CHEMOTAXIS pvals_CEU bonf_CEU fdr_CEU pvals_YRI bonf_YRI fdr_YRI 1 0.000379 0.0649 0.0324 0.4730 1 0.691 2 0.000207 0.0355 0.0324 0.6850 1 0.831 3 0.002506 0.4285 0.1012 0.2540 1 0.447 4 0.002959 0.5060 0.1012 0.3830 1 0.583 5 0.002764 0.4727 0.1012 0.8360 1 0.929 6 0.013680 1.0000 0.1533 0.0268 1 0.112

Results:

• Within-pathway: 4 (CEU) and 10 (YRI) non-redundant significant coevolution signals discovered
• Between-pathway: 16 (CEU) and 59 (YRI) significant coevolution signals discovered
• Primarily immune-associated themes

SUMMARY * Proof-of-concept pipeline successfully discovers instances of coevolution among pathways influencing population fitness * Pathways demonstrate population-driven signals of positive selection (FST) between Europeans and Africans * LD maintains inter-chromosomal SNP-SNP interactions within and between pathways enriched for positive selection

BaderLab/POPPATHR documentation built on Dec. 17, 2021, 9:53 a.m.