Blind predictions of protein interfaces by docking calculations in CAPRI
Why this work is in the frame
A frame that forgets how it found something cannot be audited. These are the routes that admitted this work.
Bibliographic record
Abstract
Reliable prediction of the amino acid residues involved in protein-protein interfaces can provide valuable insight into protein function, and inform mutagenesis studies, and drug design applications. A fast-growing number of methods are being proposed for predicting protein interfaces, using structural information, energetic criteria, or sequence conservation or by integrating multiple criteria and approaches. Overall however, their performance remains limited, especially when applied to nonobligate protein complexes, where the individual components are also stable on their own. Here, we evaluate interface predictions derived from protein-protein docking calculations. To this end we measure the overlap between the interfaces in models of protein complexes submitted by 76 participants in CAPRI (Critical Assessment of Predicted Interactions) and those of 46 observed interfaces in 20 CAPRI targets corresponding to nonobligate complexes. Our evaluation considers multiple models for each target interface, submitted by different participants, using a variety of docking methods. Although this results in a substantial variability in the prediction performance across participants and targets, clear trends emerge. Docking methods that perform best in our evaluation predict interfaces with average recall and precision levels of about 60%, for a small majority (60%) of the analyzed interfaces. These levels are significantly higher than those obtained for nonobligate complexes by most extant interface prediction methods. We find furthermore that a sizable fraction (24%) of the interfaces in models ranked as incorrect in the CAPRI assessment are actually correctly predicted (recall and precision ≥50%), and that these models contribute to 70% of the correct docking-based interface predictions overall. Our analysis proves that docking methods are much more successful in identifying interfaces than in predicting complexes, and suggests that these methods have an excellent potential of addressing the interface prediction challenge.
Fetched live from OpenAlex and de-inverted. Abstracts are not stored in this database: the inverted indexes are 8.6 GB of the frame’s 9.3 GB of text, and the host has 13 GB free.
Full frame distilled prediction
Teacher imitationNot calibrated prevalence, not ground truth. Human validation pending. Learned from the 10,348 direct Codex labels and 10,348 direct Gemma labels. Candidate is the union of thresholded teacher heads; consensus is their intersection. These outputs are machine_predicted_unvalidated and are not human labels or direct frontier model labels.
Codex and Gemma teacher scores by category
| Category | Codex | Gemma |
|---|---|---|
| Metaresearch | 0.000 | 0.000 |
| Meta-epidemiology (narrow) | 0.000 | 0.000 |
| Meta-epidemiology (broad) | 0.000 | 0.000 |
| Bibliometrics | 0.000 | 0.000 |
| Science and technology studies | 0.000 | 0.000 |
| Scholarly communication | 0.000 | 0.000 |
| Open science | 0.000 | 0.000 |
| Research integrity | 0.000 | 0.000 |
| Insufficient payload (model declined to judge) | 0.000 | 0.000 |
Machine scores (provisional)
The two teacher heads of the student model, read on this work. A score orders the frame for review; it never asserts a category, and the validation status ships verbatim with every row.
Baseline scores from an immature model (maturity gate not passed, 7 training rounds). Scores rank; they never assert a category.
score_only:v0-immature-baseline · verbatim from the scoring run: score_only means the number may rank works, and no category label ships from it