The evolution of maize yields under water deficit conditions is of particular concern in the context of climate change and human population growth. However, the multiplicity and versatility of drought-response mechanisms make the design of new drought-tolerant varieties a complex task that would greatly benefit from a better understanding of the genotype-phenotype relationship. To provide novel insight into this relationship, we aimed to build a multiscale network that best explained the genetic variance of the genotype x water deficit interaction of six drought-responsive ecophysiological traits (biomass, leaf area, water use, water use efficiency, transpiration rate, and stomatal conductance). To address this issue, we implemented an original systems biology method combining genome-wide association study, network inference, and statistical modeling on a multi-omic dataset, including phenotypic, proteomic, and genomic data acquired from 254 maize hybrids grown under well-watered (WW) and water deficit (WD) conditions. This dataset was supplemented with plasticity indices calculated as the WD/WW log ratios for protein abundances and WD/WW ratios for phenotypic traits. Using our method, we were able to build a multiscale network that increased the explained portion of the genotype x water availability interaction variance of ecophysiological traits by 20 points as compared to a classical GWAS (84% vs. 65%). This network included 63 proteins linked to six ecophysiological traits, and 531 unique loci associated, whether directly or indirectly, with ecophysiological traits. The proteins in the network were involved in signaling, protein folding, and oxidation-reduction processes. Of the loci involved in the network, three were located in genomic regions detected only by the protein analysis and contributed to more than 7 points of the total part of the GxW variance captured. These genomic regions provide a list of candidate genes, several of which can physically interact with the proteins involved in the network. Overall, our results show that multi-omics data integration can be an efficient way to capture missing heritability for complex phenotypic traits and identify new candidate genes related to drought response.