Back

K-MARVEL: K-Mer based Antimicrobial Resistance Virtual Exploration Lab

Mahar, N. S.; Branders, S.; Grabherr, M.; Gupta, I.; Ahmad, R.

2025-10-10 bioinformatics
10.1101/2025.10.08.681127 bioRxiv
Show abstract

The rapid global spread of antimicrobial resistance (AMR) necessitates a new generation of computational tools for its surveillance. While next-generation sequencing offers unprecedented insight into the resistome, current methods face a trade-off: assembly-based approaches are computationally expensive and struggle with complex metagenomes, whereas direct-mapping of long reads is hampered by high error rates that obscure critical resistance-conferring mutations. Here, we present K-MARVEL (K-Mer based Antimicrobial Resistance Virtual Exploration Lab), a novel, open-source method to capture ARGs and resistance-conferring mutations from short and long-read sequencing datasets. It operates in protein k-mer space, providing inherent tolerance to nucleotide-level sequencing errors. On a comprehensive benchmark of 61 long and 49 short-read diverse datasets, K-MARVEL demonstrated superior accuracy, achieving F1-scores of 0.9783 and 0.9754 for short and long-read datasets, respectively. Its implementation in Rust enables high speed through parallelization while guaranteeing memory safety. Computationally, it demonstrated superior performance to conventional assembly-based methods, achieving an average speed up of 7x on short-read datasets and 5x on long-read datasets. In terms of memory footprint, it outperformed the assembly-based approaches for short-read datasets, but its memory footprint was comparable for long-read datasets. Notably, K-MARVEL accurately reconstructs functional genes from genomically fragmented evidence, providing a more comprehensive resistome assessment. In conclusion, K-MARVEL provides a scalable, flexible and memory-efficient solution for AMR surveillance. Its unique capabilities for handling noisy long-read data and complex genomic scenarios make it a powerful tool for researchers and public health scientists. K-MARVEL is open-source and freely available at https://bitbucket.org/amr-avenger/k-marvel under the GPL version 3 license.

Matching journals

The top 6 journals account for 50% of the predicted probability mass.

50% of probability mass above

"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.