Figure 6.

SCs identified by FASST within the β-lactamases. Applying FASST to expose the substructural diversity of a catalytic substructure among the β-lactamases reveals many distinct clusters within the family. The GMM clustering step of FASST identifies 13 sub-groups within the family and the colors/shapes of points in the SCs correspond to cluster assignment. MESH then constructs one consensus motif for each cluster identified, resulting in an ensemble of 13 motifs. Function prediction sensitivity improves from 35.0% (single-structure motif) to 81.2% when using the motif ensemble constructed by FASST-MESH. For the highly diverse family of β-lactamases, the SCs output by FASST shows that many distinct sub-groups exist within the family. MESH takes advantage of this information to more completely model the geometric diversity present, thereby improving functional annotation coverage of the family. Mapping Family- and Phylum-level phylogenetic data to each of the substructures as shown in the corresponding plots on the right reveals that some, but not all, of the clusters identified are due to evolutionary distance between proteins. For example, the Bacillaceae proteins can be seen to form a single sub-group while Enterobacteriaceae proteins are distributed throughout the SCs in several clusters, indicating that another biological factor is working in concert with phylogenetic distance among the family of β-lactamases to produce the structural diversity uncovered by FASST.