TIGP (BIO)—Extracting Potential Antimicrobial Resistance Biomarker Genes Using Bacterial Pan-Genome-Based Feature Selection Methods

Abstract

Antimicrobial resistance is becoming a serious problem in current medical environments. Even though antibiotic drugs are still the mainstream cure for most infectious diseases, bacteria also evolved resistances against the drugs in the treatment process. The MEGA-plate experiment perfectly showed that bacteria can evolved strong resistances against antibiotic drugs very quickly and easily, suggesting that we need to pay more attention to the underlying process of antimicrobial resistances in order to fight against infectious diseases.

In this talk I will introduce how we deal with the antimicrobial resistance problem using bacterial pan-genome-based machine learning feature selection approach. Since genes are central to the antimicrobial resistance mechanisms, I will introduce how we utilize the gene-based pan-genome to catalog the bacterial gene content. I will then talk about how we apply machine learning feature selection algorithm on the gene catalog to 1) identify genes high-relevant to the resistance patterns, and 2) enhance prediction of antimicrobial resistance pathogens. I will also introduce our novel feature selection approach, the Cross-Validated Feature Selection algorithm, in extracting the minimum set of genes that may serve as antimicrobial resistance gene biomarkers in the prediction process. I hope through this talk one can learn how genes can be cataloged using pan-genomes and how machine learning feature selection algorithms can be applied on the resistance prediction problems.

附件下載