Figure 1.

Schematic of a simple graph-based algorithm for constructing a library of structural templates for homology modeling. For each connected component in the graph of sequences, where an edge represents the ability to homology model one sequence based on another, we employ a greedy approach to find a good library of template structures that cover as much of the sequence space as possible. For computation of the sequence set $\left\{{\mathcal{L}}_{\mathrm{proj}}\right\}$ of interest for experimental characterization, we skip consideration of the structures and run the algorithm on the subset with structure-associated sequences removed.