Text-mining-based feature selection for anticancer drug response prediction.

Area of Science:

• Bioinformatics
• Computational Biology
• Genomics

Background:

• Predicting anticancer treatment response from genomic data is crucial for personalized medicine.
• Machine learning models are widely used for predicting drug response from gene expression data.
• Identifying relevant features (genes) from large datasets is a significant challenge in model construction.

Purpose of the Study:

• To evaluate the efficacy of text-mining extracted features for machine learning models in predicting anticancer drug response.
• To compare text-mining based feature selection with traditional methods.
• To assess the generalizability of models trained on in vitro data to in vivo cancer models.

Main Methods:

• Utilized genes extracted via text-mining of scientific literature as features.
• Applied machine learning models to two independent cancer pharmacogenomic datasets.
• Compared performance against traditional feature selection techniques.

Main Results:

• Text-mining-based features significantly outperformed traditional feature selection methods.
• Machine learning models utilizing text-mining features demonstrated strong predictive performance.
• Models trained on in vitro data successfully predicted responses in in vivo cancer models.

Conclusions:

• Text-mining offers an effective and easily implementable approach for feature selection in machine learning for anticancer drug response prediction.
• This method enhances the development of personalized medicine strategies.
• The approach shows promise for bridging in vitro and in vivo predictive modeling.