Phylogenetic surveillance of the HIV epidemic amongst Men having Sex with Men (MSM) has revealed that large transmission networks (20+ infections/ cluster) rose from 13% of new infections in 2004 to 49% of infections in 2016 in Quebec. Identifying and responding to these ‘active’ transmission hubs in close to real-time will have the greatest impact in controlling the epidemic.
First genotypes were obtained from treatment-naïve MSM (n=4029) and heterosexual/intravenous drug user (IDU) (n=1072) groups having subtype B HIV-1 infections, as well as non-B subtype groups (n=1248). Unique non-nominative patient identifiers were assigned based on putative cluster group association, ascertained by Maximum likelihood (ML) methodology (high bootstrap support >97% and short genetic distance <0.015). Growth trajectories dynamics of 40 individual large transmission networks (20+ members/cluster) were compared with the San Diego-based HIV-TRACE (Transmission Cluster Engine) platform, and the Montreal-based Gap (distance-based) BD-SIR (cluster birth death), and the DM-PhyClus (Bayesian-based) methodologies.
Heat maps indicated overlap between estimates produced by seven clustering algorithms, revealing the role of large transmission networks in the growth of the provincial epidemic in MSM. Cluster inferences with HIV-TRACE and DM-Phys were rapid, conducive to real-time monitoring of cluster dynamics. In general, putative cluster assignments by HIV-TRACE designated at <1.5% TN93 genetic distance measures paralleled ML-based assigned. Problematic issues arose in resolving and deducing transmission links of individual members within clusters using repeat patient sampling. HIV-TRACE could not resolve two K103N and WT waves for cluster 99 and several different non-B subtypes coalesced.
In this study, we compared the sensitivity and accuracy of different phylogenetic based methodologies in estimating transmission linkage and mapping epidemic growth in close to real-time. While several cluster-based algorithms can identify ‘actively’ growing transmission hubs, resolving the linkage of individual members within clusters will require further optimization to maximize accuracy.