Requence: Genomic AMR Prediction (LaaS)

Antimicrobial resistance is projected to claim up to 10 million lives annually by 2050, hitting low- and middle-income countries hardest. Requence attacks that timeline: where traditional susceptibility testing takes three to five days, it delivers explainable resistance predictions in four to six hours.

How the pipeline works

A pathogen FASTA file enters a Lab-as-a-Service architecture that identifies the organism and routes it to a dedicated prediction endpoint. From 900+ genome assemblies we engineered more than 10,000 features (amino-acid k-mers at k=10, SNPs, targeted resistance genes), feeding gradient boosted classifiers per outcome class.

• Pearl: per-antibiotic resistance probability scores with SHAP-based attribution (e.g. +0.45 gyrA mutation impact)
• Harmony: anonymized surveillance aggregation connecting local cases to global monitoring
• Consensus checks across full and partial models guard against unreliable inference

Currently supporting Escherichia coli and Salmonella enterica.