Identification of Genomic Signatures for Colorectal Cancer Survival Using Exploratory Data Mining.

Area of Science:

• Oncology
• Genomics
• Artificial Intelligence

Background:

• Clinicopathological features are crucial for colorectal cancer (CRC) treatment but have limited prognostic value in Stage 2.
• Accurate prediction of tumor recurrence is essential for optimizing adjuvant therapy decisions in CRC.

Purpose of the Study:

• To develop and validate a novel artificial intelligence (AI)-based genomic signature for predicting recurrence in Stage 2 and 3 CRC patients.
• To integrate mutational and clinical data for enhanced molecular stratification and prognostic accuracy in CRC.

Main Methods:

• Utilized an explainable artificial intelligence (XAI) algorithm to analyze TCGA data (n=378) from Stage 2/3 CRC patients.
• Integrated genomic data from the FoundationOne Companion Diagnostic (F1CDx) assay with clinical features.
• Developed and validated a 32-gene genomic signature for predicting tumor recurrence.

Main Results:

• A 32-gene genomic signature demonstrated high precision in predicting tumor recurrence in Stage 2 CRC patients.
• Validation on an independent dataset (n=149) confirmed high prognostic accuracy (AUC: 0.952, PPV: 0.974, NPV: 0.923).
• The identified genomic signatures correlate with clinical presentation, aiding in subgroup identification.

Conclusions:

• The developed AI-driven 32-gene signature significantly improves recurrence prediction accuracy in Stage 2 CRC.
• This genomic signature holds potential to enhance molecular stratification alongside NCCN guidelines for personalized CRC management.
• The findings support the use of AI and genomic data for more precise prognostic assessments in colorectal cancer.