Building a translational cancer dependency map for The Cancer Genome Atlas.

Area of Science:

• Oncology
• Computational Biology
• Genomics

Background:

• Cancer dependency maps reveal tumor vulnerabilities for drug targeting.
• The Cancer Genome Atlas (TCGA) lacks dependency maps for patient tumors.
• Translating gene essentiality in patient tumors is crucial for therapeutic development.

Purpose of the Study:

• To develop translational dependency maps for patient tumors using machine learning.
• To identify tumor vulnerabilities that predict drug responses and disease outcomes.
• To map gene tolerability in healthy tissues for prioritizing therapeutic windows.

Main Methods:

• Applied machine learning to construct translational dependency maps from TCGA data.
• Integrated gene essentiality and tolerability data to identify vulnerabilities.
• Experimentally validated synthetic lethalities in vitro and in vivo.

Main Results:

• Identified tumor vulnerabilities predicting drug responses and patient survival.
• Discovered patient-translatable synthetic lethalities, including PAPSS1/PAPSS12 and CNOT7/CNOT78.
• Validated PAPSS1 synthetic lethality, linked to PTEN deletion and patient survival.

Conclusions:

• Translational dependency maps are valuable tools for cancer drug discovery.
• Machine learning can identify novel therapeutic targets and predict treatment outcomes.
• A web application is available for exploring tumor vulnerabilities.