ScAI: an unsupervised approach for the integrative analysis of parallel single-cell transcriptomic and epigenomic profiles

Suoqin Jin,Qing Nie,Qing Nie,Qing Nie,Lihua Zhang

doi:10.1186/s13059-020-1932-8

Abstract

Simultaneous measurements of transcriptomic and epigenomic profiles in the same individual cells provide an unprecedented opportunity to understand cell fates. However, effective approaches for the integrative analysis of such data are lacking. Here, we present a single-cell aggregation and integration (scAI) method to deconvolute cellular heterogeneity from parallel transcriptomic and epigenomic profiles. Through iterative learning, scAI aggregates sparse epigenomic signals in similar cells learned in an unsupervised manner, allowing coherent fusion with transcriptomic measurements. Simulation studies and applications to three real datasets demonstrate its capability of dissecting cellular heterogeneity within both transcriptomic and epigenomic layers and understanding transcriptional regulatory mechanisms.

Highlights

The rapid development of single-cell technologies allows for dissecting cellular heterogeneity more comprehensively at an unprecedented resolution
Using the normalized scRNA-seq data matrix X1 (p genes in n cells) and the single-cell chromatin accessibility or DNA methylation data matrix X2 (q loci in n cells) as an example, we infer the low-dimensional representations via the following matrix factorization model: minW 1;W 2;H;Z ≥ 0αkX1 þ þ γkXX2ðZ∘HRÞ:−jW21; 2H
By performing hierarchical clustering of the calculated deviations of top 30 most variable transcription factors (TFs), we found that these TFs were divided into 2 clusters, and each TF cluster was specific to 1 particular cell subpopulation, which was found to be consistent with the clustering by single-cell aggregation and integration (scAI) (Additional file 2: Figure S5)

Summary

Introduction

The rapid development of single-cell technologies allows for dissecting cellular heterogeneity more comprehensively at an unprecedented resolution. Many protocols have been developed to quantify transcriptome [1], such as CEL-seq, Smart-seq, Drop-seq, and 10X Chromium, and techniques that measure single-cell chromatin accessibility (scATAC-seq) and DNA methylation have become available [2]. Several single-cell multiomics technologies have emerged for measuring multiple types of molecules in the same individual cell, such as scM&T-seq [3], scNMT-seq [4], scTrio-seq [5], sci-CAR-seq [6], and scCAT-seq [7]. The resulting singlecell multiomics data has potential of providing new insights regarding the multiple regulatory layers that control cellular heterogeneity [8, 9].

Results

Discussion

Conclusion