Pathotypr: harmonised MTBC lineage assignment and resistance-associated variant detection for genomic surveillance

Abstract BACKGROUND Rapid, interoperable whole-genome tools for Mycobacterium tuberculosis complex (MTBC) surveillance remain limited for harmonised lineage assignment across recognised lineages and simultaneous resistance-associated variant detection. AIM To develop and validate Pathotypr, an alignment-free tool for harmonised MTBC lineage assignment and resistance genotyping from assemblies and raw reads. METHODS We reconstructed an MTBC phylogeny from 26,813 genomes using 609,003 polymorphic sites, derived an updated lineage marker backbone, and implemented a k-mer/Random Forest framework with marker-based lineage and WHO catalogue-based resistance calling. Performance was evaluated on 498 RefSeq assemblies, 88,071 UShER-TB typed sequencing samples, 162 clinical read sets for closest-reference matching, and 7,148 CRyPTIC isolates with phenotypic drug susceptibility data. RESULTS Pathotypr supported all 14 currently recognised MTBC lineages (L1-L10, A1-A4). On 498 complete genomes, marker-based and alignment-free lineage calls were 100% concordant, and prediction accuracy remained 100% on 254 independent assemblies. In 88,071 non-ambiguous UShER-TB samples, root-lineage concordance with TB-Profiler was 100%, while Pathotypr additionally identified lineage 10, A1 and A2. Resistance predictions showed 85.0% genotype-phenotype concordance overall, with high performance for rifampicin (95.8% sensitivity, 95.0% specificity) and isoniazid (93.0%, 97.9%). Runtime was about 1 second per sample, enabling analysis of 88,071 samples in approximately 24 hours on four threads. In the MDR-enriched CRyPTIC collection, Pathotypr supported reconstruction of 135 probable introduction events into Germany, Italy and Ukraine; 33.7% of introduction-associated isolates carried MDR/pre-XDR genotypes. CONCLUSION Pathotypr enables rapid, harmonised MTBC lineage assignment and high-confidence resistance screening, supporting near real-time and cross-border tuberculosis surveillance.