Known QAC determinants outperform genome language model embeddings in a leakage-aware public Listeria benzalkonium chloride tolerance benchmark.

Known QAC determinants achieve a balanced accuracy of 0.961 in predicting benzalkonium chloride tolerance in 197 Listeria monocytogenes isolates, outperforming genome language models.

• Why it matters: Accurate phenotype prediction from genomic data is crucial for understanding disinfectant resistance, but current models often lack reliable, isolate-linked labels and proper evaluation methods, limiting their practical use.
• What they did: The study compared traditional sequence features, DNABERT2-derived DNA representations, and known QAC resistance determinants across multiple data splits, assessing their predictive performance on a public benchmark.
• The result: Known QAC determinants proved most effective, enabling high-confidence predictions and guiding future research, though larger, linked datasets are needed to validate and extend these findings.