A machine learning approach for ranking clusters of docked protein‐protein complexes by pairwise cluster comparison

Abstract

Reliable identification of near‐native poses of docked protein–protein complexes is still an unsolved problem. The intrinsic heterogeneity of protein–protein interactions is challenging for traditional biophysical or knowledge based potentials and the identification of many false positive binding sites is not unusual. Often, ranking protocols are based on initial clustering of docked poses followed by the application of an energy function to rank each cluster according to its lowest energy member. Here, we present an approach of cluster ranking based not only on one molecular descriptor (e.g., an energy function) but also employing a large number of descriptors that are integrated in a machine learning model, whereby, an extremely randomized tree classifier based on 109 molecular descriptors is trained. The protocol is based on first locally enriching clusters with additional poses, the clusters are then characterized using features describing the distribution of molecular descriptors within the cluster, which are combined into a pairwise cluster comparison model to discriminate near‐native from incorrect clusters. The results show that our approach is able to identify clusters containing near‐native protein–protein complexes. In addition, we present an analysis of the descriptors with respect to their power to discriminate near native from incorrect clusters and how data transformations and recursive feature elimination can improve the ranking performance. Proteins 2017; 85:528–543. © 2016 Wiley Periodicals, Inc.

Protein docking refinement by convex underestimation in the low-dimensional subspace of encounter complexes

“…We will next employ a machine learning approach for ranking the 30 clusters generated from the SSDU set. Some related work on using machine learning approaches, different than ours, for ranking has appeared in the literature 46 , 47 . We used several classification algorithms on this dataset: random forests, support vector machines with linear and radial kernels and logistic regression.…”

SKEMPI 2.0: An updated benchmark of changes in protein-protein binding energy, kinetics and thermodynamics upon mutation

“…SKEMPI is a manually curated database of mutations in structurally characterised protein-protein interactions and the effect of those mutations on binding affinity and other parameters [47]. The first release has been used as a basis for many further studies, including the development of energy functions [48], [46] which were subsequently implemented in the CCharPPI web server for characterising protein-protein interactions [49], as well as being used for ranking docked poses [45], [58], [6], [50]. SKEMPI has also been used to study human disease [56], [16], [55], assessing the role of dynamics on binding [69], exploring the conservation of binding regions [28], evaluating experimental affinity measurement methods [22], as well serving as a data source for models which predict dissociation rate changes upon mutation [1], pathological mutations [23], hotspot residues (e.g.…”

Predicting improved protein conformations with a temporal deep recurrent neural network

“…The results show that the proposed RNN model with GRU cells is able to outperform classical machine learning methods such as RF, LR and KNN, which are representative of state-of-the-art classical machine learning algorithms and have been successfully applied to other bioinformatic domains [26][27][28]. In particular, the model presented here achieves a mean precision of 0.415 on the validation set of the CV compared to 0.121, 0.140 and 0.000 for KNN, RF and LR, respectively.…”