Table 3.
The table shows the average Kendall's τ and average Pearson's r using benchmark B. Correlation is computed separately for each target, then averaged. The Align-all row is the results of MQA with distance constraints from alignments using the 6 constraint sets described here. The Align-only row is the results of MQA with no noncontacts. The Meta-weighted and meta-unweighted rows are the results of extracting constraints from the models to be assessed (with weighted models and unweighted models respectively). TASSER, Lee, and Pcons are top-ranked MQA methods presented at CASP7 (groups 125, 556, and 634 respectively). Qiu is a newer MQA method described in Qiu et al. [15]. The companion paper by Archie and Karplus [3], evaluates our MQA algorithm on more measures.
| Group | τ̅ | r̄ |
|---|---|---|
| Meta-weighted | 0.624 | 0.862 |
| Meta-unweighted | 0.624 | 0.861 |
| Lee | 0.585 | 0.805 |
| Qiu | 0.581 | 0.853 |
| Align-all | 0.574 | 0.832 |
| Align-only | 0.570 | 0.832 |
| Pcons | 0.560 | 0.847 |
| TASSER | 0.538 | 0.633 |