Workflow chart of mixture modeling approach. The figure depicts a general scheme of our mixture modeling method. First, one selects the query gene(s) from a biological process of interest. The respective synthetic lethal data set, retrieved either from a database, such as the GRID, or from own data then defines the list of target genes to consider . Integration of different genomic information sources and generation of genomic features, that characterize the relationship of the query to its synthetic lethal targets, results in a multivariate data set of pairwise scores. Application of a Gaussian Mixture Model identifies a small group of target genes. Varying the posterior probability of the model refines the partitioning, such that the small group shows significant enrichment with known genes in the biological process of the query gene(s), if possible. As a last step, follow-up screens characterize the candidate genes contained in the small group in the biological context that is given (e.g. involvement in spindle migration).