Precise prediction of multiple anticancer drug efficacy using multi target regression and support vector regression...

Area of Science:

• Oncology
• Computational Biology
• Bioinformatics

Background:

• Precision medicine in oncology aims to predict anti-cancer drug efficacy using patient-specific tumor attributes.
• Oral squamous cell carcinoma (OSCC) presents a challenge for selecting effective therapeutic strategies.

Purpose of the Study:

• To develop and validate machine learning models for predicting the efficacy of five anti-cancer drugs in OSCC.
• To computationally rank potential anti-cancer drugs based on predicted efficacy using tumor characteristics.

Main Methods:

• Machine learning models, including multi-target regression (MTR) and support vector regression (SVR), were developed.
• Novel pre-processing techniques enhanced existing models to create E_MTR, EL_MTR, EM_SVR, and ELM_SVR.
• Models were trained and cross-validated using both real OSCC samples and theoretical samples.

Main Results:

• MTR models showed lower error than SVR models with 30 real tumor samples.
• With 340 theoretical samples, SVR models achieved near-zero error, outperforming MTR.
• Enhanced models demonstrated accurate drug efficacy prediction and ranking, matching actual rankings.

Conclusions:

• Developed statistical and machine learning models (MTR and SVR) for predicting anti-cancer drug efficacy.
• Enhanced models, particularly EL_MTR and ELM_SVR, offer improved accuracy and precise drug ranking for personalized oncology.
• These models support precision medicine by enabling the selection of the most suitable anti-cancer drugs for individual patients.