Single-cell genomic technologies provide an unprecedented opportunity to define molecular cell types in a data-driven fashion, but present unique data integration challenges. Here we present UINMF, a computational tool designed to improve dataset integrations by leveraging unshared features.