Genome-wide scale analyses identify novel BMI genotype-environment interactions using a conditional false discovery rate

Abstract

Genotype-environment interaction (G×E) studies typically focus on variants with previously known marginal associations. While such two-step filtering greatly reduces the multiple testing burden, it can miss loci with pronounced G×E effects, which tend to have weaker marginal associations. To test for G×E effects on a genome-wide scale whilst leveraging information from marginal associations in a flexible manner, we combine the conditional false discovery rate with interaction test results obtained from StructLMM. After validating our approach, we applied this strategy to UK Biobank (UKBB) data to probe for G×E effects on BMI. Using 126,077 UKBB individuals for discovery, we identified known (FTO, MC4R, SEC16B) and novel G×E signals, many of which replicated (FAM150B/ALKAL2,TMEM18, EFR3B, FAIM2, UNC79, LAT) in an independent subset of UKBB (n=126,076). Finally, when analysing the full UKBB cohort, we identified 140 candidate loci with G×E effects, highlighting the advantages of our approach.

Optimising expression quantitative trait locus mapping workflows for single-cell studies

“…To this end, we selected a recently proposed multiple testing correction method that conditions the false discovery rate (i.e. conditional FDR, or cFDR) on an external set of test statistics [39,40] and tested the impact of its application to sc-eQTL mapping (Fig. 5a).…”

“…We also tested alternatives to the Storey Q value for the second step. In particular, we test 1) the Bonferroni approach (p-value<0.05), which controls the much stricter family-wise error rate (FWER), 2) Benjamini-Hochberg (BH), another commonly used FDR approach;and 3) a recent implementation [39,40] of the conditional FDR (cFDR), which leverages external data to guide the gene-level multiple testing correction. For the cFDR procedure we tested using raw association p-values from 1) the bulk iPSC associations, 2) GTEx v7 association p-values from the EBV transformed lymphocytes, 3) GTEx v7 association p-values from tissue meta-analyses results [55], and 4) association p-values from a joint eQTL mapping on the mesendoderm and endoderm data from the Cuomo et al study (Cuomo et al 2020).…”