Library-Based Microbial Source Tracking via Strain Identification

Microbial Source Tracking (MST) aims to classify the source host-species of biological matter, typically fecal matter, using strains of fecal indicator bacteria, often E. coli. This paper continues addressing the MST problem using analysis of a library of bacterial fingerprints started in [9]. The Cal Poly Library of Pyroprints (CPLOP) is a collection of fingerprints of over 6,000 E. coli isolates collected from the fecal matter of a variety of host-species. In prior work [9] we studied the accuracy of the MST process based on k-Nearest Neighbors discovery in CPLOP. This process, while sufficiently accurate, does not scale well with the size of the database. In this paper, we study the accuracy of a clustering-based MST approach which scales significantly better: the bacterial isolate information stored in CPLOP is clustered using an efficient density-based clustering technique. We present our analysis of the accuracy and efficiency of the clustering-based MST methodology for CPLOP.