Finalist
Team 13319: Jimena Martin Reina (Team lead, Universidad Carlos III de Madrid), Robert Beardmore (University of Exeter), Pablo Catalan Fernandez (Universidad Carlos III de Madrid), and Isabel Garcia Jimenez (Universidad Carlos III de Madrid)
Project Overview. Antimicrobial resistance (AMR) is an evolving landscape that requires equally dynamic surveillance tools. Our multidisciplinary team bridging microbiology, mathematics, and biotechnology proposes a rigorous analysis of the ATLAS database to decode the underlying structure of resistance. With preliminary analyses already completed, we are very well positioned to deliver actionable insights within the challenge’s 8-week timeframe.
Data & Scope. We will focus on the rich, raw Minimum Inhibitory Concentration (MIC) data for Enterobacteriaceae (specifically E. coli, K. pneumoniae, and E. cloacae) within the ATLAS dataset. Covering thousands of isolates across 20 antibiotics from 2018 onwards, this longitudinal data allows for a high-resolution study of resistance trends.
Our approach is twofold: 1) Topological Mapping of Resistance Profiles: We will treat resistance phenotypes as binary vectors to construct a similarity-based network. By visualizing these nodes across temporal and geographic scales, we can identify the “evolutionary trajectories” of resistance. This reveals dominant clusters, emergent threats, and the velocity at which novel resistance profiles spread, allowing for the early identification of high-risk strains.
2) Algorithmic Optimization of Surveillance Panels: Current surveillance is often resource-heavy. Using Principal Component Analysis (PCA) and greedy algorithms, we will identify linear dependencies within the 20-antibiotic panel. Our preliminary findings suggest the effective dimensionality of this data is significantly lower than current testing protocols. By defining a “minimal viable panel” that retains maximal information, we can drastically reduce the cost and time required for global surveillance.
Impact. This project moves beyond static reporting. By combining topological data analysis with dimensionality reduction, our work translates raw MIC data into actionable guidance for a faster and more cost-effective AMR surveillance system.