Genotype–phenotype associations

We used the DGRP Freeze 2.0 webtool for genome-wide association analysis (dgrp2.gnets.ncsu.edu). This tool tested sleep latency against 1,920,276 genome-wide polymorphisms segregating in the DGRP having minor allele frequencies ≥ 0.05 (9 or more lines having the minor allele)¹⁷. The webtool first adjusts phenotypes for cryptic genetic relatedness, the presence of chromosomal inversions, and Wolbachia pipientis infection status¹⁷. Neither the inversions nor Wolbachia infection status was significantly associated with sleep latency; however, adjustments were made to the sleep latency phenotype based on cryptic relatedness (Supplementary Table S1). Using the adjusted data, we then applied the following linear mixed model:

where y is the adjusted sleep latency means, X is the design matrix containing the fixed effect of each SNP, Z is the incidence matrix of random effects, and e is the random residual error¹⁷.

We defined our threshold P-value for each association as P ≤ 1 × 10^–5, a threshold used in many other studies of the DGRP^{22,37,41,90–94} and supported by Q-Q plots (Supplementary Fig. S1). Variants meeting the threshold criterion were mapped to the Drosophila 6.0 genome. We used the DRSC Integrative Ortholog Prediction Tool (DIOPT) to identify Drosophila genes having human homologs²⁷, and the Database for Annotation, Visualization, and Integrated Discovery (DAVID) to assess gene ontology and pathway enrichment²⁷.