Explainable Machine Learning Identifies Factors for Dosage Compensation in Aneuploid Human Cancer Cells

Area of Science:

• Genetics
• Molecular Biology
• Cancer Biology

Background:

• Aneuploidy, common in cancer, alters protein levels due to changes in chromosome copy number.
• Despite widespread aneuploidy, protein levels are often buffered towards diploid levels, but mechanisms are unclear.

Purpose of the Study:

• To quantify protein dosage buffering in cancer.
• To identify molecular mechanisms and predictors of protein dosage compensation.
• To explore the adaptive advantage and therapeutic relevance of protein dosage compensation in aneuploid cancers.

Main Methods:

• Developed a novel approach to quantify protein dosage buffering.
• Utilized multifactorial machine learning models to identify buffering predictors.
• Analyzed protein dosage compensation in cancer cell lines and in vivo tumor samples.

Main Results:

• Protein dosage compensation is widespread but variable in aneuploid cancer samples.
• Key predictors of buffering include gene dependency, protein complex participation, haploinsufficiency, and mRNA decay.
• Higher buffering correlates with reduced proteotoxic stress and increased drug resistance, impacting oncogenic potential.

Conclusions:

• Protein dosage compensation is a significant regulatory mechanism in aneuploid cancers.
• Understanding buffering predictors can inform therapeutic strategies.
• Protein dosage compensation represents a potential therapeutic target for cancer treatment.