SCCA: a contrastive learning framework to capture shared features and target-specific variations in single-cell RNA-seq data

Abstract

Single-cell RNA sequencing (scRNA-seq) is widely used to investigate differences between experimental and biological conditions, such as the effects of gene knockouts or drug treatments. Recently, contrastive analysis, designed to identify variations enriched in a target data set compared to a control data set, has gained popularity and shown considerable success in these studies. These approaches aim to disentangle the common variation present in both datasets from the salient variation that is unique to the target dataset. However, existing contrastive analysis methods often lack the sensitivity to detect subtle variations. In this work, we introduce SCCA, a novel method that integrates contrastive learning with an iterative clustering objective to yield disentangled latent representations that are more sensitive to small salient effects. We evaluated the performance of SCCA by comparing it with six state-of-the-art algorithms across various single-cell RNA sequencing datasets. Our results demonstrate that SCCA consistently outperforms competing methods in identifying salient effects and achieving well-disentangled latent representations, thereby enhancing both the sensitivity and interpretability of single-cell analyses.

Keywords

contrastive learning / contrastive analysis / single cell RNA-seq / cluster

RIGHTS & PERMISSIONS

Higher Education Press 2026

Just Accepted

This article has successfully passed peer review and final editorial review, and will soon enter typesetting, proofreading and other publishing processes. The currently displayed version is the accepted final manuscript. The officially published version will be updated with format, DOI and citation information upon launch. We recommend that you pay attention to subsequent journal notifications and preferentially cite the officially published version. Thank you for your support and cooperation.