SPN-PBP-AMR

Analysis of the Streptococcus pneumoniae PBPs and inferred MICs.

PBP Analysis and Predicted MICs

This Streptococcus pneumoniae task carries out an analysis of the pbp1A, pbp2B and pbp2X genes, firstly assigning an allele code to each one and then using a machine learning approach to estimate the MICs (minimum inhibitory concentration) for a set of antimicrobials. The resistance phenotype is then interpreted using CLSI guidelines, including where there are different thresholds for the meningital and non-meningital forms. In testing, the MICs were predicted to >97% correct (to one dilution) and resistance categorisation was >93%.

The initial development of the machine learning model is described in Li et al (2016) while the method is benchmarked in Li et al (2017). The software and databases for this task have been provided by Ben Metcalf at the Centre for Disease Control (CDC) and is available from https://github.com/BenJamesMetcalf/Spn_Scripts_Reference