AUTHOR=Yu Jianwei , Jia Yan , Yu Qichao , Lin Lan , Li Chao , Chen Bowang , Zhong Pingyu , Lin Xueqing , Li Huilan , Sun Yinping , Zhong Xuejing , He Yuqi , Huang Xiaoyun , Lin Shuangming , Pan Yuanming 

TITLE=Deciphering complex antibiotic resistance patterns in Helicobacter pylori through whole genome sequencing and machine learning

JOURNAL=Frontiers in Cellular and Infection Microbiology

VOLUME=Volume 13 - 2023

YEAR=2024

URL=https://www.frontiersin.org/journals/cellular-and-infection-microbiology/articles/10.3389/fcimb.2023.1306368

DOI=10.3389/fcimb.2023.1306368

ISSN=2235-2988

ABSTRACT=Helicobacter pylori (H.pylori, Hp) affects billions of people worldwide. Despite the availability of different antibiotics, emerging resistance of Hp renders antibiotic treatment ineffective. Next generation sequencing provides a powerful technology to investigate the genotype-phenotype connection for Hp. However, the prediction of antibiotic resistance using whole genome sequencing data remains a formidable challenge. Here we analyzed 52 Hp strains from different hospitals and assessed their susceptibility to five antibiotics using the agar dilution assay. Whole-genome equencing was performed to screen the anti-microbial resistance (AMR) genotypes of the Hp strains. The relationship between drug resistance and genotype was modeled with univariate statistical tests, unsupervised machine learning and supervised machine learning. We established support vector machine models for Amoxicillin resistance (66% sensitivity and 100% specificity) and Clarithromycin resistance (100% sensitivity and 100% specificity), which is better than the known resistance site of the 23S rRNA gene for Amoxicillin (A1834G, 22.2% sensitivity and 100% specificity) and Clarithromycin (A2147, 87% sensitivity and 96% specificity). Our results proved that predictive modeling using supervised learning algorithm with feature selection could lead to diagnostic model with higher predictive power as compared with models relying on single SNP site. Our study contributes valuable insights towards enhancing precision and effectiveness in antibiotic treatment strategies for Hp infections with the application of whole-genome sequencing for Hp.