Table 1. Runtime and sensitivity analyses.
| Structural Motif | Hits | Time | First FN | R.M.S.D. | FNR <1 Å |
|---|---|---|---|---|---|
| serine protease [2] | 3,498 | 0.92 s | 1hcg | 0.72 Å | 1.86% |
| aminopeptidase [3] | 350 | 0.46 s | N/A | N/A | 0.00% |
| zinc coordination [5] | 1,056 | 0.13 s | 2eou | 0.42 Å | 8.52% |
| enolase superfamily [6] | 288 | 0.36 s | 3mkc | 0.89 Å | 2.33% |
| enolase (exchanges) | 308 | 0.87 s | 3mkc | 0.89 Å | 5.62% |
| RNA G-tetrad [7] | 84 | 1.10 s | 2rsk | 1.53 Å | 0.00% |
Each reported runtime represents the average over 10 benchmark runs. Results are compared to an exhaustive search strategy [27] and the minimum R.M.S.D. of any false negative (FN) hit is reported. The false negative rate (FNR) below 1 Å (i.e., a general cutoff below which we consider hits to be biologically meaningful) is given. S1 Table shows that false negative rate can be reduced at the cost of a moderate runtime increase.