Antimicrobial resistance genetic factor identification from whole-genome sequence data using deep feature selection.

Area of Science:

• Genomics
• Bioinformatics
• Machine Learning

Background:

• Antimicrobial resistance (AMR) poses a significant global health threat, necessitating the understanding of its biological mechanisms.
• Whole-genome single nucleotide polymorphism (SNP) data combined with AMR profiles offers a powerful resource for identifying AMR-associated mutations.
• Machine learning, particularly feature selection, can leverage genomic data to detect genetic factors contributing to AMR.

Purpose of the Study:

• To develop and apply a supervised feature selection approach using deep neural networks to identify AMR-associated genetic factors from whole-genome SNP data.
• To detect AMR-associated mutations in Neisseria gonorrhoeae using whole-genome sequence data.

Main Methods:

• The study employed a deep neural network-based method called DNP-AAP (deep neural pursuit - average activation potential).
• DNP-AAP utilized whole-genome SNP data and AMR profiles for feature selection.
• Logistic regression classifiers were built using the identified SNPs to predict antibiotic resistance.

Main Results:

• DNP-AAP successfully identified known AMR-associated genes in Neisseria gonorrhoeae.
• The method generated a list of candidate SNPs potentially linked to novel AMR determinants.
• High prediction AUCs (0.949-0.994) were achieved for resistance to penicillin, tetracycline, azithromycin, ciprofloxacin, and cefixime.

Conclusions:

• DNP-AAP effectively identifies known AMR-associated genes and suggests novel AMR factors in Neisseria gonorrhoeae.
• The DNP-AAP method is broadly applicable to AMR analysis across bacterial species with genomic and phenotype data.
• This tool can aid microbiologists in screening for genetic candidates for experimental validation.