Evaluating the Performance of Sputum-Based Targeted Sequencing Against Mycobacterial Whole-Genome Sequencing in Predicting Tuberculosis Drug Resistance

Direct sputum targeted next-generation sequencing (tNGS) offers rapid resistance profiling without culture, but its concordance with isolate-based whole-genome sequencing (WGS) and phenotypic drug susceptibility testing (pDST) remains unclear. We compared tNGS (direct sputum) and WGS (cultured isolates) for 14 drugs' resistance prediction in 68 culture-positive tuberculosis specimens, using pDST as reference. Performance for lineage concordance was also assessed. tNGS showed the highest rifampicin (RIF) sensitivity (90.9%) but the lowest specificity (65.2%); WGS achieved the best overall agreement (86.8%). For isoniazid, tNGS sensitivity was 82.6% vs. WGS 78.3%, but WGS specificity was higher (91.1% vs. 75.6%). tNGS outperformed WGS for ethambutol (EMB) sensitivity (80.0% vs. 40.0%). Both methods performed poorly for pyrazinamide (PZA) (agreement ~40%). Among 68 specimens, 51.5% had fully concordant resistance profiles; tNGS-only variants outnumbered WGS-only variants 2:1. Crucially, tNGS and WGS on the same 24 cultured isolates yielded identical results, proving that discrepancies arise from culture-driven clonal selection, not technical differences. Direct sputum tNGS suggests broader within-host resistance diversity that may be missed by culture, whereas WGS offers superior specificity. The two methods are complementary, with culture bias being the primary source of discordance.