The premise underlying the proposed Stream Greedy Construction scheme in [28] is that instead of re-evaluating all the k diverse results for each sliding window, the diverse subset of the current window is initialized using the diverse results from the previous window.
The challenge of computing diverse subsets for each query (Q_i) in ({mathbb {M}}) is that those diverse subsets cannot be computed incrementally.
Having just moved my analytic office to a new location, patients' associations to the new office were explored and had significant and diverse meaning for each patient.
Therefore, a selection of diverse conformers for each compound is considered for the purposes of 3-D neighboring.
Note that we used only the first diverse conformer for each compound, being the PubChem default conformer.
Using the first diverse conformer for each compound, the ST values for these 13.2 billion pairs were computed using ROCS [20] from OpenEye software, Inc., consuming ~419 CPU days in total, and stored.
In the first part of this study, the 3-D similarity score distribution curves were generated for the unique conformer-conformer, conformer-compound, and compound-compound pairs, using a single conformer and ten diverse conformers for each of 10,000 randomly selected compounds (Figures 1, 2, 3, 4 and Table 3).
For every combination of size and type, we simulated three reference databases, for a total of 6 (random only) or 12 (random and diverse) databases for each gene family.
However, we presented alignments of all RAMP families (for these alignments, we tried to select the most diverse representatives for each family) discussed in this work in Additional File 1 to back all our claims.
