CATFISHR: integration of transcriptome and inferred CNV profiles for comprehensive malignant cell identification

CATFISHR integrates transcriptome similarity with inferred CNV and consensus typing to identify malignant cells, achieving balanced accuracy > 0.85 in six of eight cohorts.

• Why it matters: Malignant cell identification in scRNA-seq is difficult because RNA typing, somatic variant detection, and inferred CNV each miss different signals, risking exclusion of rare malignant populations.
• What they did: CATFISHR is a semi-automated workflow evaluated on eight cohorts (seven scRNA-seq datasets across multiple carcinomas and melanoma subtypes, plus one 10x Visium spatial transcriptomics sample) using benchmarks or ground truths.
• The result: It achieved F1 score > 0.80 in six of eight cohorts, and its “Hidden” malignant calls appeared in every dataset, including Visium spots with histologist-confirmed morphologically malignant cells.