BERMUDA: a novel deep transfer learning method for single-cell RNA sequencing batch correction reveals hidden high-resolution cellular subtypes

Zixiao Lu,Bryan R. Helm,Jie Zhang,Kun Huang,Tongxin Wang,Travis S. Johnson,Wei Shao

doi:10.1186/s13059-019-1764-6

Abstract

To fully utilize the power of single-cell RNA sequencing (scRNA-seq) technologies for identifying cell lineages and bona fide transcriptional signals, it is necessary to combine data from multiple experiments. We present BERMUDA (Batch Effect ReMoval Using Deep Autoencoders), a novel transfer-learning-based method for batch effect correction in scRNA-seq data. BERMUDA effectively combines different batches of scRNA-seq data with vastly different cell population compositions and amplifies biological signals by transferring information among batches. We demonstrate that BERMUDA outperforms existing methods for removing batch effects and distinguishing cell types in multiple simulated and real scRNA-seq datasets.

Highlights

Single-cell transcriptional dynamics are important for understanding molecular physiology and disease dysregulation within heterogeneous tissues
Framework of BERMUDA We propose BERMUDA, a novel unsupervised framework to remove batch effects across different batches by training an autoencoder (Fig. 1a)
In order to successfully remove batch effects, we propose a novel approach by combining the reconstruction loss with the transfer loss when training the autoencoder

Summary

Introduction

Single-cell transcriptional dynamics are important for understanding molecular physiology and disease dysregulation within heterogeneous tissues. The standard techniques for single-cell analysis were flow cytometry [1, 2] and fluorescence imaging of tissue on slides [3, 4]. Though these techniques have provided tremendous insights, they are limited to a small, pre-defined set of molecular markers [1]. Singlecell RNA sequencing (scRNA-seq) was developed to characterize high-throughput gene expression profiles for populations of individual cells, which has enabled an unprecedented resolution of cellular heterogeneity in complex tissues. Widespread adoption of scRNA-seq techniques has produced large complex datasets, which

Methods

Results

Discussion

Conclusion