While only recently developed, the ability to profile expression data in single cells (scRNA-Seq) has already led to several important studies and findings. However, this technology has also raised several new computational challenges. These include questions about the best methods for clustering scRNA-Seq data, how to identify unique group of cells in such experiments, and how to determine the state or function of specific cells based on their expression profile. To address these issues researchers at Carnegie Mellon University develop and test a method based on neural networks (NN) for the analysis and retrieval of single cell RNA-Seq data. They tested various NN architectures, some of which incorporate prior biological knowledge, and used these to obtain a reduced dimension representation of the single cell expression data. They show that the NN method improves upon prior methods in both, the ability to correctly group cells in experiments not used in the training and the ability to correctly infer cell type or state by querying a database of tens of thousands of single cell profiles. Such database queries (which can be performed using their web server) will enable researchers to better characterize cells when analyzing heterogeneous scRNA-Seq samples.
Network architectures of NN models used in this paper
(A) Single-layered fully connected network. (B) Two-layered fully connected network. (C) Single-layered network with TF and PPI cluster nodes (connected only to their member genes) and fully connected dense nodes. (D) Two-layered network that is similar to the model in (C) but with an additional fully connected layer. Note that bias nodes are also included in each model.
Availability – Supporting website: http://sb.cs.cmu.edu/scnn/.