A leakage-aware genomic prediction pipeline for meropenem resistance in Klebsiella pneumoniae using transformer-based resistome representation learning

İlkay Sibel Kervancı, Marion Adebiyi · 2026

Read A leakage-aware genomic prediction pipeline for meropenem resistance in Klebsiella pneumoniae using transformer-based resistome representation learning by İlkay Sibel Kervancı, Marion Adebiyi (2026) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/PMC13532757.txt.

Description

Kervancı and Adebiyi built in 2026 a prediction pipeline that reads bacterial genomes to forecast meropenem resistance. Their target was Klebsiella pneumoniae, a hospital bacterium often resistant to this last-line antibiotic. Many machine learning models flatter themselves by leaking family links between training and test bacteria, and flat gene lists miss gene-to-gene interplay. Their pipeline splits data by bacterial lineage to block that leak, and uses a transformer encoder, an attention-based model, to turn gene presence patterns into rich features.

Kervancı found the hybrid system beat standard methods such as logistic regression and random forests. The best setup reached accuracy above nine in ten on internal tests, and checks showed most of the gain came from the learned gene features rather than the final classifier. On an outside set of 305 samples from several countries, the model still held up well, which supports use across settings.

Catalog record

CREATORİlkay Sibel Kervancı, Marion Adebiyi
YEAR2026
COLLECTIONPAPERS
FORMATTXT
FILE SIZE49 KB
LICENSEhttps://creativecommons.org/licenses/by/4.0
RIGHTSCC BY
SOURCEEurope PMC

Source: https://europepmc.org/article/MED/PMC13532757

Browse the full library · About the Tower of Babel