Most phenotypes show complex inheritance and continuous variation, which play a central role in evolution and susceptibility to disease. We have developed a suite of computational methods for the elucidation of the genetic contribution to molecular, cellular, and organismal traits. Of particular importance in this context are epistatic interactions, i.e. the unexpected effects of allele combinations from different loci. For example, by exploiting the information contained in known family trios we could for the first time detect allele incompatibilities in a mouse population. Based on data from a mouse panel with more than 2,000 heterozygous mice we detected 168 statistically significant interactions, which is substantially more than what had previously been reported in the literature. Further, we developed extensions of the Random Forest machine learning method, enabling us to detect epistatic interactions in QTL mapping data with unprecedented sensitivity. E.g. by applying this framework to yeast proteomics data we found that almost one third of all QTL were involved in epistatic effects on protein concentrations, underlining the importance of epistasis for understanding complex traits. Finally, the application of these ideas to RNA-sequencing data has revealed the importance of the non-coding transcriptome for understanding the molecular mechanisms of phenotype variation.