Machine learning-driven multi-targeted drug discovery in colon cancer using biomarker signatures

Area of Science:

• Computational oncology
• Bioinformatics
• Artificial Intelligence in Medicine

Background:

• Challenges persist in understanding colon cancer (CC) molecular pathways and identifying essential genes for targeted therapies.
• Leveraging molecular data, such as gene expression and mutation profiles, is crucial for advancing multi-targeted therapies in CC.
• Current therapeutic strategies require improvement in predicting drug responses and addressing drug resistance.

Purpose of the Study:

• To develop and validate an integrated computational framework for personalized colon cancer (CC) therapy.
• To enhance the prediction of therapeutic outcomes and drug responses using advanced AI algorithms.
• To address challenges in CC treatment, including drug resistance and the need for precision medicine.

Main Methods:

• Integration of biomarker signatures from high-dimensional gene expression, mutation data, and protein interaction networks.
• Application of Adaptive Bacterial Foraging (ABF) optimization for refining search parameters and maximizing predictive accuracy.
• Utilization of the CatBoost algorithm for patient classification based on molecular profiles and drug response prediction.

Main Results:

• The proposed ABF-CatBoost system demonstrated superior performance over traditional Machine Learning models, achieving 98.6% accuracy, 0.984 specificity, 0.979 sensitivity, and 0.978 F1-score.
• The model accurately predicts toxicity risks, metabolism pathways, and drug efficacy profiles, contributing to safer and more effective treatments.
• External validation confirmed the model's predictive accuracy and generalizability for personalized colon cancer therapy.

Conclusions:

• The developed AI model significantly improves precision medicine in colon cancer (CC) by personalizing therapy based on patient-specific molecular profiles.
• This computational framework offers a multi-targeted therapeutic approach, addressing drug resistance and optimizing drug selection and dosage.
• The adaptable nature of the framework allows for its modification for other cancer types, expanding its impact on personalized cancer treatment and accelerating drug discovery.