Exploring Prediction of Antimicrobial Resistance Based on Protein Solvent Accessibility Variation

Area of Science:

• Computational biology
• Genomics
• Drug discovery

Background:

• Antimicrobial resistance (AMR) is a major global health concern.
• Accurate identification of resistant bacterial strains is crucial for effective AMR surveillance and treatment.
• Existing AMR detection methods struggle to identify novel resistance genes or variants not present in reference databases.

Purpose of the Study:

• To investigate the relationship between protein variants associated with AMR and their relative solvent accessibility (RSA).
• To develop a novel computational scoring system, RSA-AMR, for predicting AMR potential based on RSA.
• To enhance the capabilities of existing AMR detection algorithms for identifying novel AMR variants.

Main Methods:

• In silico analysis of protein variants and their correlation with AMR.
• Development of the RSA-AMR scoring system based on the association between AMR and RSA.
• Integration of the RSA-AMR score with existing AMR detection algorithms.

Main Results:

• Known AMR protein variants are significantly associated with exposed residues, while susceptible variants are often buried.
• The RSA-AMR scoring system effectively estimates a protein variant's propensity to alter RSA and influence AMR.
• Integrating RSA-AMR with existing algorithms improved specificity by tenfold in detecting novel AMR variants.

Conclusions:

• RSA-AMR offers a novel computational approach to predict AMR potential based on protein structure.
• This scoring system can augment current AMR detection tools, expanding their scope to novel variants.
• Further development is needed to address single-point changes and expand the model learning dataset for broader applicability.