22The quest for genome-wide signatures of selection in populations using SNP data has 23 proven efficient to uncover genes involved in conserved or adaptive molecular functions, but 24 none of the statistical methods were designed to identify interacting genes as targets of 25 selective processes. Here, we propose a straightforward statistical test aimed at detecting 26 epistatic selection, based on a linkage disequilibrium (LD) measure accounting for population 27 structure and heterogeneous relatedness between individuals. SNP-based ( ) and window-28 based ( 1 ) statistics fit a Student distribution, allowing to easily and quickly test the 29 significance of correlation coefficients in the frame of Genome-Wide Epistatic Selection 30 Scans (GWESS) using candidate genes as baits. As a proof of concept, use of SNP data from 31 the Medicago truncatula symbiotic legume plant uncovered a previously unknown gene 32 coadaptation between the MtSUNN (Super Numeric Nodule) receptor and the MtCLE02 33 (CLAVATA3-Like) signalling peptide, and experimental evidence accordingly supported a 34 MtSUNN-dependent negative role of MtCLE02 in symbiotic root nodulation. Using human 35 HGDP-CEPH SNP data, our new statistical test uncovered strong LD between SLC24A5 and 36 EDAR worldwide, which persists after correction for population structure and relatedness in 37 Central South Asian populations. This result suggests adaptive genetic interaction or 38 coselection between skin pigmentation and the ectodysplasin pathway involved in the 39 development of ectodermal organs (hairs, teeth, sweat glands), in some human populations. 40 Applying this approach to genome-wide SNP data will foster the identification of 41 evolutionary coadapted gene networks. 42 43 Author summary 46 Population genomic methods have allowed to identify many genes associated with 47 adaptive processes in populations with complex histories. However, they are not designed to 48 identify gene coadaptation between genes through epistatic selection, in structured 49 populations. To tackle this problem, we developed a straightforward LD-based statistical test 50 accounting for population structure and heterogeneous relatedness between individuals, using 51 SNP-based ( ) or windows-based ( 1 ) statistics. This allows easily and quickly testing 52 for significance of correlation coefficients between polymorphic loci in the frame of Genome 53 Wide Epistatic Selection Scans (GWESS). Following detection of gene coadaptation using 54 SNP data from human and the model plant Medicago truncatula, we report experimental 55 evidence of genetic interaction between two receptors involved in the regulation of root 56 nodule symbiosis in Medicago truncatula. This test opens new avenues for exploring the 57 evolution of genes as interacting units and thus paves the way to infer new networks based on 58 evolutionary coadaptation between genes. 59 60 65 selection against deleterious mutations (1-3). Identifying genes showing such selection 66 signatures has been a major goal of population...