Single-cell RNA sequencing (scRNA-seq) data are noisy and sparse. Here, UPENN researchers show that transfer learning across datasets remarkably improves data quality. By coupling a deep autoencoder with a Bayesian model, SAVER-X extracts transferable gene-gene relationships across data from different labs, varying conditions and divergent species, to denoise new target datasets.

Outline of the SAVER-X transfer learning framework

a, The autoencoder pretraining step. b, Workflow of SAVER-X. For target data with a UMI count matrix, SAVER-X trains the target data with autoencoder without a chosen pretraining model (item A), then filters unpredictable genes using cross-validation (item B) and estimates the final denoised values with empirical Bayesian shrinkage (item C).

Availability – SAVER-X is publicly available at http://singlecell.wharton.upenn.edu/saver-x/