P21

Oxford Nanopore Technologies (ONT) could decentralize Mycobacterium tuberculosis (Mtb) sequencing but the need for high (≥50 ng) DNA concentrations and suboptimal accuracy prevented its implementation. We assessed the performance of ONT R10 flowcells and V14 rapid barcoding chemistry for Mtb whole genome sequencing of DNA extracted from clinical primary liquid cultures (CPLCs). Using the recommended protocol for MinION Mk1C, R10.4.1 MinION flowcells and ONT Rapid Barcoding Kit V14 on six CPLC samples, the pooled library yield was 10.8 ng/ul, 25Gb of data and 214k reads were generated after 48h, but 49% of reads failed to meet the Phred Quality score threshold (>8). Because we hypothesized that the poor performance was due to impurities in CPLCs, we added a pre-library preparation bead-cleanup step. The yield for four CPLCs and one Mtb subculture (control) was similar (10.8 ng/ul), and 32Gb of data and 822k reads were produced but the quality remained poor (66% reads with Phred Quality >8). A Phred Quality score >20 (modal accuracy above 99%) was only achieved for samples that underwent bead-cleanup.  A third run of five CPLCs with bead clean-up only produced 12Gb of data, 166k reads of which only 51% achieved Phred Quality score >8. Using TB-profiler, a median depth of coverage above 10x was only achieved in four of 17 sequenced libraries and drug resistance profiles could not be determined. These results suggest that further optimization of the latest ONT rapid barcoding chemistry and library preparation protocol is needed before nanopore sequencing can guide tuberculosis care.