Using Neural Networks To Improve Single-Cell RNA-Seq Data Analysis

摘要: While only recently developed, the ability to profile expression data in single cells (scRNA-Seq) has already led several important studies and findings. However, this technology also raised new computational challenges including questions related handling noisy sometimes incomplete data, how identify unique group of such experiments determine state or function specific based on their profile. To address these issues we develop test a method neural networks (NN) for analysis retrieval cell RNA-Seq data. We tested various NN architectures, some biologically motivated, used obtain reduced dimension representation show that improves upon prior methods both, correctly not training infer type by querying database tens thousands profiles. Such queries (which can be performed using our web server) will enable researchers better characterize when analyzing heterogeneous scRNA-Seq samples.