A framework for target discovery in rare cancers.

Area of Science:

• Genomics
• Oncology
• Computational Biology

Background:

• Large-scale functional genetic screens have identified cancer dependencies, but rare cancers are underrepresented.
• The landscape of gene dependencies in many rare cancers remains largely unknown.
• TFE3-translocation renal cell carcinoma (tRCC) is a rare cancer with limited research on its genetic dependencies.

Purpose of the Study:

• To identify novel cancer dependencies in rare cancers, specifically TFE3-translocation renal cell carcinoma (tRCC).
• To develop and apply a machine learning model to predict gene dependencies in rare cancers lacking experimental models.
• To nominate actionable vulnerabilities in poorly-characterized cancer types.

Main Methods:

• Genome-scale CRISPR knockout screens were performed in tRCC models.
• Machine learning models were trained to infer gene dependencies from transcriptional profiles.
• Dependency prediction was applied to alveolar soft part sarcoma (ASPS) and a large cohort of TCGA tumors and other rare cancers.

Main Results:

• CRISPR screens in tRCC revealed dependencies in mitochondrial biogenesis, oxidative metabolism, and kidney lineage specification pathways.
• Machine learning successfully predicted gene dependencies, identifying MCL1 as a dependency in ASPS but not tRCC.
• The predictive model identified potential vulnerabilities across multiple rare cancers, including 13 subtypes of kidney cancer.

Conclusions:

• Functional genetic screening coupled with predictive modeling can establish a landscape of candidate vulnerabilities in rare cancers.
• This approach can uncover previously unknown, cancer-selective dependencies, such as MCL1 in ASPS.
• The findings provide a foundation for developing targeted therapies for rare cancers with limited treatment options.