Evaluating an automated clustering approach in a perspective of ongoing surveillance of porcine reproductive and respiratory syndrome virus (PRRSV) field strains.

Porcine reproductive and respiratory syndrome virus (PRRSV) has a major economic impact on the swine industry. The important genetic diversity needs to be considered for disease management. In this regard, information on the circulating endemic strains and their dispersal patterns through ongoing surveillance is beneficial. The objective of this project was to classify Quebec PRRSV ORF5 sequences in genetic clusters and evaluate stability of clustering results over a three-year period using an in-house automated clustering system. Phylogeny based on maximum likelihood (ML) was first inferred on 3661 sequences collected in 1998-2013 (Run 1). Then, sequences collected between January 2014 and September 2016 were sequentially added into 11 consecutive runs, each one covering a three-month period. For each run, detection of clusters, which were defined as groups of ≥15 sequences having a≥70% rapid bootstrap support (RBS) value, was automated in Python. Cluster stability was described for each cluster and run based on the number of sequences, RBS value, maximum pairwise distance and agreement in sequence assignment to a specific cluster. First and last run identified 29 and 33 clusters, respectively. In the last run, about 77% of the sequences were classified by the system. Most clusters were stable through time, with sequences attributed to one cluster in Run 1 staying in the same cluster for the 11 remaining runs. However, some initial groups were further subdivided into subgroups with time, which is important for monitoring since one specific wild-type cluster increased from 0% in 2007 to 45% of all sequences in 2016. This automated classification system will be integrated into ongoing surveillance activities, to facilitate communication and decision-making for stakeholders of the swine industry.