Online Single-cell RNA-seq Data Denoising with Transfer Learning
Abstract
Single-cell RNA sequencing (scRNA-seq) technology brings in unprecedented opportunities to new findings in fields including immunology, neuroscience and cancer research. However, the data is still very noisy and suffers from low capture rates. We develop an open-to-public gateway where users can perform online data denoising to improve the quality of their single-cell RNA sequencing datasets. Our gateway can provide a free, convenient, fast and reliable parallel computation platform to handle more than 50K cells at one time. The gateway is based on SAVER-X, a computational and statistical tool that combines deep autoencoder with Bayesian inference for scRNA-seq denoising, and features transfer learning from relevant public datasets. It allows general users and clinicians to improve their data quality without seeking additional computational resources or statistical training, thus would benefit researchers with a wide range of backgrounds.
