SMNN: Batch Effect Correction for Single-cell RNA-seq data via Supervised Mutual Nearest Neighbor Detection

SMNN is a flexible, accurate and robust method for batch effect correction of scRNA-seq data. SMNN performs batch effect correction via supervised mutual nearest neighbor detection. Specifically, in our current implementation, SMNN either takes cluster/cell-type label information as input or infers cell types using scRNA-seq clustering. It then detects mutual nearest neighbors within matched cell types and corrects batch effect accordingly. Compared to the unsupervised methods, SMNN provides improved merging within the corresponding cell types across batches, and retains more cell type specific features after correction, especially under realistic scenarios where different batches may easily differ in many aspects including samples used, single cell capture technology employed, or library preparation approach adopted. For details, please refer to our tutorial.

Comments and suggestions are welcome, please e-mail Yuchen Yang at yyuchen@email.unc.edu or Yun Li at yunli@med.unc.edu.