Identifying functionally significant microRNAs (miRs) and their correspondingly most important messenger RNA targets (mRNAs) in specific biological contexts is a critical task to improve our understanding of molecular mechanisms underlying organismal development, physiology and disease. However, current miR–mRNA target prediction platforms rank miR targets based on estimated strength of physical interactions and lack the ability to rank interactants as a function of their potential to impact a given biological system. To address this, we have developed ToppMiR (http://toppmir.cchmc.org), a web-based analytical workbench that allows miRs and mRNAs to be co-analyzed via biologically centered approaches in which gene function associated annotations are used to train a machine learning-based analysis engine. ToppMiR learns about biological contexts based on gene associated information from expression data or from a user-specified set of genes that relate to context-relevant knowledge or hypotheses. Within the biological framework established by the genes in the training set, its associated information content is then used to calculate a features association matrix composed of biological functions, protein interactions and other features. This scoring matrix is then used to jointly rank both the test/candidate miRs and mRNAs. Results of these analyses are provided as downloadable tables or network file formats usable in Cytoscape.