Understanding the relationship between genetic variation and phenotypic variation for quantitative traits is necessary for predicting responses to natural and artificial selection and disease risk in human populations, but is challenging because of large sample sizes required to detect and validate loci with small effects. Here, we used the inbred, sequenced, wild-derived lines of the Drosophila melanogaster Genetic Reference Panel (DGRP) to perform three complementary genome-wide association (GWA) studies for natural variation in olfactory behavior. The first GWA focused on single nucleotide polymorphisms (SNPs) associated with mean differences in olfactory behavior in the DGRP, the second was an extreme quantitative trait locus GWA on an outbred advanced intercross population derived from extreme DGRP lines, and the third was for SNPs affecting the variance among DGRP lines. No individual SNP in any analysis was associated with variation in olfactory behavior by using a strict threshold accounting for multiple tests, and no SNP overlapped among the analyses. However, combining the top SNPs from all three analyses revealed a statistically enriched network of genes involved in cellular signaling and neural development. We used mutational and gene expression analyses to validate both candidate genes and network connectivity at a high rate. The lack of replication between the GWA analyses, small marginal SNP effects, and convergence on common cellular networks were likely attributable to epistasis. These results suggest that fully understanding the genotype-phenotype relationship requires a paradigm shift from a focus on single SNPs to pathway associations.genetic architecture | chemosensation | behavioral genetics U nderstanding the rules by which variation in primary DNA sequence impacts variation for quantitative traits in natural populations is critical for predicting responses to natural and artificial selection and disease risk in human populations. The emerging picture from large genome-wide association (GWA) studies in human populations is that many common variants with individually small marginal (additive) effects affect diseases and quantitative traits, of which only a small fraction can be replicated across populations (1, 2). Although it is possible that effects of common single nucleotide polymorphisms (SNPs) underestimate the true effects because causal variants are not common or not SNPs, and lack of replication is due to differences in allele frequency and pattern of linkage disequilibrium (LD) (1, 2), it is also possible that small additive effects and lack of replication are due to underlying epistatic interactions (3, 4).Extremely large samples are required to determine individual significance of rare alleles and epistatic interactions. One approach to gaining biological insight from GWA studies in the absence of statistical significance of individual SNPs is functional evaluation of genes harboring the top SNPs regardless of individual significance. A second approach is to consider the genes w...