A predictive algorithm for the analysis of AMR trends and healthcare decision support

Abstract

Background Translating available AMR surveillance data to observe evolving patterns of microbial resistance to antimicrobial agents while identifying regions at high risk of resistant cases and serving as a decision-support tool is an aspect of AMR surveillance that is rarely explored nationwide and uncommon globally. Therefore, we developed a two-tier dashboard algorithm (PATHFINDER) that can analyse antimicrobial surveillance datasets, observe evolving global spatiotemporal patterns of AMR, and integrate local AMR gene resources to identify functional AMR determinant genes and antibiotic classes from the query organism genome. Methods The Python-based plotly library was used to develop the interactive variables of the Antimicrobial Testing Leadership Surveillance (ATLAS) dataset in an adjustable spatiotemporal environment. A lightweight database containing multiple known resistant genes from the ResFinder database was used as a prototype to identify unique AMR determinant genes from query nucleotide sequences. In R, a function was created to accept query genome sequences and generate Kmers of length 250 using the blaster package. The GPT-4 API plug-in was embedded with adequate prompt parsing for it as an interpretation LLM function. Results AMR trendline plots were designed for invasive infections and customised based on the class of antibiotics and infection types on the surveillance dashboard. The decision-support tool correctly predicted resistant genes with a sensitivity of 75% on pre-confirmed organisms. The observed specificity score (51.5%) was due to the need for more filtering and optimisation and not to PATHFINDER performance. When run against the reference gene dataset containing pre-identified AMR genes, the support tool generated a BLAST table with identified AMR gene determinants in a nucleotide sequence. Conclusions The PATHFINDER algorithm has the potential to revolutionise healthcare decision-making. It can inform targeted interventions, guide antimicrobial stewardship efforts at a national level, promote appropriate antibiotic use, and significantly reduce the risk of resistance development.