Optimizing nanopore adaptive sampling for pneumococcal serotype surveillance in complex samples using the graph-based GNASTy algorithm

Serotype surveillance of <i>Streptococcus pneumoniae</i> (the pneumococcus) is critical for understanding the effectiveness of current vaccination strategies. However, existing methods for serotyping are limited in their ability to identify co-carriage of multiple pneumococci and detect novel serotypes. To develop a scalable and portable serotyping method that overcomes these challenges, we employed nanopore adaptive sampling (NAS), an on-sequencer enrichment method that selects for target DNA in real-time, for direct detection of <i>S. pneumoniae</i> in complex samples. Whereas NAS targeting the whole <i>S. pneumoniae</i> genome was ineffective in the presence of nonpathogenic streptococci, the method was both specific and sensitive when targeting the capsular biosynthetic locus (CBL), the operon that determines <i>S. pneumoniae</i> serotype. NAS significantly improved coverage and yield of the CBL relative to sequencing without NAS and accurately quantified the relative prevalence of serotypes in samples representing co-carriage. To maximize the sensitivity of NAS to detect novel serotypes, we developed and benchmarked a new pangenome-graph algorithm, named GNASTy. We show that GNASTy outperforms the current NAS implementation, which is based on linear genome alignment, when a sample contains a serotype absent from the database of targeted sequences. The methods developed in this work provide an improved approach for novel serotype discovery and routine <i>S. pneumoniae</i> surveillance that is fast, accurate, and feasible in low-resource settings. Although NAS facilitates whole-genome enrichment under ideal circumstances, GNASTy enables targeted enrichment to optimize serotype surveillance in complex samples.