Predicting clinical outcomes using cancer progression associated signatures

Area of Science:

• Genomics
• Cancer Biology
• Computational Biology

Background:

• Somatic mutation signatures offer insights into cancer causes but have limited predictive power for patient outcomes.
• Understanding cancer progression requires novel predictive biomarkers.

Purpose of the Study:

• To assess the effectiveness of 142 mutation-signature-associated metrics (P142) for predicting cancer progression.
• To evaluate the utility of these metrics in a large TCGA PanCancer Atlas cohort.

Main Methods:

• Utilized mutation data from 10,437 patients across 31 cancer types from The Cancer Genome Atlas (TCGA).
• Developed cancer-specific machine learning models (XGBoost) using P142 metrics to predict Progression Free Survival (PFS) status (High vs. Low).
• Employed stratified random sampling for training, tuning, and validation cohorts, with performance evaluated on the validation set.

Main Results:

• The P142 panel successfully predicted PFS status in specific cancer types, including adrenocortical carcinoma, glioma, mesothelioma, and sarcoma.
• Model performance varied across different cancer types, indicating differential utility of the P142 metrics.
• The study identified specific mutation signatures associated with cancer progression.

Conclusions:

• The P142 panel demonstrates potential as a predictive tool for cancer progression in select malignancies.
• These findings highlight the importance of mutation signatures in understanding cancer aetiology and patient prognosis.
• Further research is warranted to refine and expand the application of these signatures in clinical settings.