scMalignantFinder distinguishes malignant cells in single-cell and spatial transcriptomics by leveraging cancer signatures

• DOI: 10.1038/s42003-025-07942-y
• Publication Date: 2025-03-27T23:34:37Z
• Authors: Qiaoni Yu; Yuan-Yuan Li; Yunqin Chen
• Keywords: Single-Cell Analysis

Single-cell RNA sequencing (scRNA-seq) is a powerful tool for characterizing tumor heterogeneity, yet accurately identifying malignant cells remains challenging. Here, we propose scMalignantFinder, machine learning specifically designed to distinguish from their normal counterparts using data- and knowledge-driven strategy. To develop the tool, multiple cancer datasets were collected, initially annotated calibrated nine carefully curated pan-cancer gene signatures, resulting in over 400,000 single-cell transcriptomes training. The union of differentially expressed genes across was taken as features model construction comprehensively capture transcriptional diversity. scMalignantFinder outperformed existing automated methods two gold-standard eleven patient-derived scRNA-seq datasets. capability predict malignancy probability empowers dynamic characteristics during progression. Furthermore, holds potential annotate regions spatial transcriptomics. Overall, provide an efficient detecting heterogeneous cell populations. Editorial Summary: identifies RNA-seq transcriptomics by leveraging signatures.