INSCT: Integrating millions of single cells using batch-aware triplet neural networks

Abstract

Efficient integration of heterogeneous and increasingly large single cell RNA sequencing (scRNA-seq) data poses a major challenge for analysis and in particular, comprehensive atlasing efforts. Here, we developed a novel deep learning algorithm to overcome batch effects using batch-aware triplet neural networks, called INSCT ("Insight"). Using simulated and real data, we demonstrate that INSCT generates an embedding space which accurately integrates cells across experiments, platforms and species. Our benchmark comparisons with current state-of-the-art scRNA-seq integration methods revealed that INSCT outperforms competing methods in scalability while achieving comparable accuracies. Moreover, using INSCT in semi-supervised mode enables users to classify unlabeled cells by projecting them into a reference collection of annotated cells. To demonstrate scalability, we applied INSCT to integrate more than 2.6 million transcriptomes from four independent studies of mouse brains in less than 1.5 hours using less than 25 gigabytes of memory. This feature empowers researchers to perform atlasing scale data integration in a typical desktop computer environment. INSCT is freely available at https://github.com/lkmklsmn/insct. HighlightsO_LIINSCT accurately integrates multiple scRNA-seq datasets C_LIO_LIINSCT accurately predicts cell types for an independent scRNA-seq dataset C_LIO_LIEfficient deep learning framework enables integration of millions of cells on a personal computer C_LI