PUREE: accurate pan-cancer tumor purity estimation from gene expression data

Area of Science:

• Oncology
• Bioinformatics
• Computational Biology

Background:

• Tumors exhibit cellular heterogeneity, with varying proportions of cancer cells (tumor purity).
• Accurate tumor purity estimation is crucial for reliable analysis of tumor gene expression data and understanding cancer biology.
• Existing methods for purity estimation have limitations, necessitating improved approaches.

Purpose of the Study:

• To develop and validate PUREE, a novel weakly supervised learning method for inferring tumor purity from gene expression profiles.
• To assess PUREE's accuracy and generalizability across diverse solid tumor types.
• To compare PUREE's performance against existing transcriptome-based purity estimation methods.

Main Methods:

• PUREE was developed using a weakly supervised learning approach.
• The model was trained on gene expression data and genomic consensus purity estimates from 7864 solid tumor samples.
• Performance was evaluated across various tumor types and validated using single-cell RNA-seq data.

Main Results:

• PUREE demonstrated high accuracy in predicting tumor purity across distinct solid tumor types.
• The method generalized well to tumor samples from previously unseen tumor types and cohorts.
• PUREE outperformed existing transcriptome-based purity estimation approaches in benchmark tests.

Conclusions:

• PUREE is a highly accurate and versatile tool for estimating tumor purity from bulk tumor gene expression data.
• This method can effectively interrogate tumor heterogeneity and complement genomics-based approaches.
• PUREE offers a valuable alternative in settings where genomic data is not available.