Updates to the integrated protein-protein interaction benchmarks: Docking benchmark version 5 and affinity benchmark version 2

By Thom Vreven, Iain H. Moal, Anna Vangone, Brian G. Pierce, Panagiotis L. Kastritis, Mieczyslaw Torchala, Raphael A. G. Chaleil, Brian Jiménez-García, Paul A. Bates, Juan Fernández-Recio and 2 more. 2015

424 citations recorded by the source corpus.

We present an updated and integrated version of our widely used protein-protein docking and binding affinity benchmarks. The benchmarks consist of non-redundant, high quality structures of protein-protein complexes along with the unbound structures of their components. Fifty-five new complexes were added to the docking benchmark, 35 of which have experimentally-measured binding affinities. These updated docking and affinity benchmarks now contain 230 and 179 entries, respectively. In particular, the number of antibody-antigen complexes has increased significantly, by 67% and 74% in the docking and affinity benchmarks, respectively. We tested previously developed docking and affinity prediction algorithms on the new cases. Considering only the top ten docking predictions per benchmark case, a prediction accuracy of 38% is achieved on all 55 cases, and up to 50% for the 32 rigid-body cases only. Predicted affinity scores are found to correlate with experimental binding energies up to r=0.52 overall, and r=0.72 for the rigid complexes.

Subfields: Molecular Biology, Radiology, Nuclear Medicine and Imaging

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

In Molecular Biology

In Radiology, Nuclear Medicine and Imaging

Citations in the corpus

Performance and its limits in rigid body protein-protein docking (2020, positive): “Success rates of the other four servers for the 51 remaining targets were adopted from the original publication (Vreven et al., 2015).”
Protein Interface Prediction using Graph Convolutional Networks (2017, methodology): “In our experiments we used the data from Version 5 of the Docking Benchmark Dataset, which is the standard benchmark dataset for assessing docking and interface prediction methods [25].”
Benchmarking AlphaFold for protein complex modeling reveals accuracy determinants (2021, methodology): “…previously assigned test case difficulty classifications, which are based on binding conformational change between unbound and bound structures, 20,23 did not markedly impact the success of AlphaFold; for acceptable or higher accuracy predictions in the set of five models, success rates for…”
P HYSICS - INFORMED DEEP NEURAL NETWORK FOR RIGID - BODY PROTEIN DOCKING (2022, background): “The PIP dataset is split such that no structure in the training set has more than 30% sequence identity with any structure in the DB5.”
Performance of HADDOCK and a simple contact-based protein–ligand binding affinity predictor in the D3R Grand Challenge 2 (2017, background): “This simple structural-based approach has led to one of the best performing predictors so far reported on a large and heterogeneous set of data [43, 44], with Pearson’s Correlation of 0.73 between the predicted and the experimental values and a root mean-squared error of 1.89 kcal mol −1 .”
Key Topics in Molecular Docking for Drug Design (2019, methodology): “There are also benchmarking databases which encompass specific complexes or purposes, such as protein-protein complexes [80], membrane protein-protein complexes [81], and a blind set based on PDBBind for testing machine learning SFs [82].”
TCRmodel: high resolution modeling of T cell receptors from sequence (2018, methodology): “Structures with resolution worse than 3.2 ˚A were removed, commensurate with the resolution cutoff used for structure selection for previously reported protein-protein docking and TCR-peptide-MHC docking benchmarks (3.25 ˚A) (17,18).”
Dockground: A comprehensive data resource for modeling of protein complexes (2018, background): “Applying more stringent criterion for the sequence identity between bound and unbound structures for both interactors (96% identity, 80% coverage, as in Weng’s benchmark 5 19 ) reduced this number to 427.”
Graph convolutional networks for computational drug development and discovery (2019, background): “Docking Benchmark Database (DBD5) [107] is a benchmark database that contains a collection of distinct protein docking test cases.”
Geometric Transformers for Protein Interface Contact Prediction (2021, methodology): “Finally, in Table 5, we observe that, in predicting the interface contacts between unbound protein chains in the DB5 test dataset, the Geometric Transformer enables enhanced top-k precision and recall (deﬁnition in A.2) compared to all other baseline methods, including GCNs and Graph Trans-formers p”