PLAGUE: Improved confidence plasmid reconstruction from short read assemblies

Abstract

As the vast majority of archival data is short-read based, reliable strategies for detecting and classifying plasmids out of these data are still needed. A widely adopted solution for binning plasmids with rich contextual data is MOB-Suite. However, a large portion of the available data has low coverage and results in fragmented assemblies. Under these circumstances, MOB-Suite delivers a high number of false positives, leading to wasteful follow-up experiments. To improve the reliability of MOB-Suite, while retaining and enriching its biological contextual output, we have developed PLAGUE. PLAGUE is a fast and low-memory consumption post-processing tool for MOB-Suite outputs, relying on checks of circularity, reads spanning the gap between the 3 and 5 ends and total coverage of the plasmid candidate. PLAGUE can reduce the number of false positives in MOB-Suite outputs by over 30%, while retaining almost full sensitivity. It is fully containerized and easy to use in clusters as well as locally.