Modified Bootstrapping and K-Means Clustering for Taxonomic Binning

Metagenomics is the study of microbial ecology using genetics as an access point. We seek to understand the microbial communities in environments such as tidal pools, soil, mine runoff, or even the human gut, so that we can understand the impact that microbes have on our world and our health. Metagenomic analysis usually involves the determination of what species are present in a given sample, and, if possible, what each species’ genome looks like. Because the vast majority of microbial organisms cannot be cultured easily in a generic lab setting, we look instead to the DNA taken from a sample, and try to assign the DNA to different species (the taxonomic binning problem). Many methods for solving the taxonomic binning problem exist, but they are extremely computationally intensive, reliant on woefully incomplete reference databases, or insufficiently accurate. In this paper, we put forward a new taxonomic binning algorithm called Horatio. Horatio is a modified bootstrapping and k-means clustering algorithm for taxonomic binning that uses compositional information about DNA subsequences. Horatio requires relatively little setup, and has tunable parameters that can reflect users’ preferences. We compare and contrast Horatio to other extant taxonomic binning algorithms and evaluate its performance against simulated and real metagenomic datasets.