I am working with mouse scRNA-seq data generated using the 10x Chromium platform and would like to distinguish tumor epithelial cells from normal epithelial cells within the same dataset.

In human datasets, I have previously used copy-number–based approaches such as inferCNV to identify malignant cells. However, I am aware that murine tumors often exhibit fewer and subtler CNV events compared to human tumors, which may limit the reliability of CNV-inference methods in mouse systems.

I am familiar with newer machine-learning–based tools such as Ikarus, scMalignantFinder, and related classifiers, but my understanding is that many of these models are trained primarily on human data and may not generalize well to mouse tumors without retraining or careful validation.

I would appreciate recommendations for well-established, widely cited, and field-accepted approaches that have been successfully used to distinguish malignant versus normal epithelial cells in mouse scRNA-seq data. In particular:

Are there CNV-based tools that are considered robust and validated for murine tumors?

Are there expression-based or pathway-level approaches that are commonly used as alternatives to CNV inference in mice?

Are there mouse-specific classifiers or reference-based methods that are broadly accepted in the literature?

Pointers to benchmark studies, best-practice reviews, or commonly used pipelines in mouse cancer scRNA-seq would be especially helpful.

Thank you for any insights or references.