P029

1:University of Oxford; 2:UK Health Security Agency (UKHSA)

Accurate identification of mycobacteria of clinical relevance, whether rare or common, and the ability to detect mixed infections is required. We are analysing competitive mapping of a large and growing number of clinical samples to develop a methodology to classify isolates into clusters, likely to represent independent species. It is, however, currently unclear how many clinical samples are required to create a near-complete catalogue.

Initially, 4443 clinical isolates of non-tuberculous mycobacteria (NTM) from Public Health England (now UKHSA) were sequenced and competitively mapped against 178 references sequences using a cloud-based pipeline. Isolates were assigned to 87 putative species, and 39 species were in clusters of at least 5 isolates (18 species >=20). For these 39 species, preliminary threshold values were established to distinguish known species, identify possible mixed infections, detect novel species and to identify the need for new reference sequences.  The results were compared with the Mykrobe species classifier containing 203 reference species. Also, the pipeline was assessed with a ‘truth’ set of sequences from 153 reference species sourced from type collections.

Ongoing evaluation of competitive mapping performance on a further >30,000, sequenced by PHE/UKHSA, is being used to confirm the thresholds for assignation of species in clusters, more precise identification of mixed infections, and recognition of new clusters (species). The most current results will be presented. This approach is designed to progressively establish a much enlarged catalogue of species and provide an estimate for the sample size needed to create a near-complete catalogue of mycobacterial species.