Many biological processes are mediated by protein-protein interactions (PPIs). Because protein domains are the building blocks of proteins, PPIs likely rely on domain-domain interactions (DDIs). Several attempts exist to infer DDIs from PPI networks but the produced datasets are heterogeneous and sometimes not accessible, while the PPI interactome data keeps growing. We describe a new computational approach called "PPIDM" (Protein-Protein Interactions Domain Miner) for inferring DDIs using multiple sources of PPIs. The approach is an extension of our previously described "CODAC" (Computational Discovery of Direct Associations using Common neighbors) method for inferring new edges in a tripartite graph. The PPIDM method has been applied to seven widely used PPI resources, using as "Gold-Standard" a set of DDIs extracted from 3D structural databases. Overall, PPIDM has produced a dataset of 84,552 non-redundant DDIs. Statistical significance (p-value) is calculated for each source of PPI and used to classify the PPIDM DDIs in Gold (9,175 DDIs), Silver (24,934 DDIs) and Bronze (50,443 DDIs) categories. Dataset comparison reveals that PPIDM has inferred from the 2017 releases of PPI sources about 46% of the DDIs present in the 2020 release of the 3did database, not counting the DDIs present in the Gold-Standard. The PPIDM dataset contains 10,229 DDIs that are consistent with more than 13,300 PPIs extracted from the IMEx database, and nearly 23,300 DDIs (27.5%) that are consistent with more than 214,000 human PPIs extracted from the STRING database. Examples of newly inferred DDIs covering more than 10 PPIs in the IMEx database are provided. Further exploitation of the PPIDM DDI reservoir includes the inventory of possible partners of a protein of interest and characterization of protein interactions at the domain level in combination with other methods. The result is publicly available at http://ppidm.loria.fr/.

An amendment to this paper has been published and can be accessed via the original article.

MOTIVATION: Protein-protein docking algorithms aim to predict the 3D structure of a binary complex using the structures of the individual proteins. This typically involves searching and scoring in a 6D space. Many docking algorithms use FFT techniques to exhaustively cover the search space and to accelerate the scoring calculation. However, FFT docking results often depend on the initial protein orientations with respect to the Fourier sampling grid. Furthermore, Fourier-transforming a physics-base force field can involve a serious loss of precision. RESULTS: Here, we present EROS-DOCK, an algorithm to rigidly dock two proteins using a series of exhaustive 3D rotational searches in which non-clashing orientations are scored using the ATTRACT coarse-grained force field model. The rotational space is represented as a quaternion 'π-ball', which is systematically sub-divided in a 'branch-and-bound' manner, allowing efficient pruning of rotations that will give steric clashes. The algorithm was tested on 173 Docking Benchmark complexes, and results were compared with those of ATTRACT and ZDOCK. According to the CAPRI quality criteria, EROS-DOCK typically gives more acceptable or medium quality solutions than ATTRACT and ZDOCK. AVAILABILITY AND IMPLEMENTATION: The EROS-DOCK program is available for download at http://erosdock.loria.fr. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

BACKGROUND: Families of related proteins and their different functions may be described systematically using common classifications and ontologies such as Pfam and GO (Gene Ontology), for example. However, many proteins consist of multiple domains, and each domain, or some combination of domains, can be responsible for a particular molecular function. Therefore, identifying which domains should be associated with a specific function is a non-trivial task. RESULTS: We describe a general approach for the computational discovery of associations between different sets of annotations by formalising the problem as a bipartite graph enrichment problem in the setting of a tripartite graph. We call this approach "CODAC" (for COmputational Discovery of Direct Associations using Common Neighbours). As one application of this approach, we describe "GODomainMiner" for associating GO terms with protein domains. We used GODomainMiner to predict GO-domain associations between each of the 3 GO ontology namespaces (MF, BP, and CC) and the Pfam, CATH, and SCOP domain classifications. Overall, GODomainMiner yields average enrichments of 15-, 41- and 25-fold GO-domain associations compared to the existing GO annotations in these 3 domain classifications, respectively. CONCLUSIONS: These associations could potentially be used to annotate many of the protein chains in the Protein Databank and protein sequences in UniProt whose domain composition is known but which currently lack GO annotation.

BACKGROUND: Many entries in the protein data bank (PDB) are annotated to show their component protein domains according to the Pfam classification, as well as their biological function through the enzyme commission (EC) numbering scheme. However, despite the fact that the biological activity of many proteins often arises from specific domain-domain and domain-ligand interactions, current on-line resources rarely provide a direct mapping from structure to function at the domain level. Since the PDB now contains many tens of thousands of protein chains, and since protein sequence databases can dwarf such numbers by orders of magnitude, there is a pressing need to develop automatic structure-function annotation tools which can operate at the domain level. RESULTS: This article presents ECDomainMiner, a novel content-based filtering approach to automatically infer associations between EC numbers and Pfam domains. ECDomainMiner finds a total of 20,728 non-redundant EC-Pfam associations with a F-measure of 0.95 with respect to a "Gold Standard" test set extracted from InterPro. Compared to the 1515 manually curated EC-Pfam associations in InterPro, ECDomainMiner infers a 13-fold increase in the number of EC-Pfam associations. CONCLUSION: These EC-Pfam associations could be used to annotate some 58,722 protein chains in the PDB which currently lack any EC annotation. The ECDomainMiner database is publicly available at http://ecdm.loria.fr/ .

MOTIVATION: Docking prediction algorithms aim to find the native conformation of a complex of proteins from knowledge of their unbound structures. They rely on a combination of sampling and scoring methods, adapted to different scales. Polynomial Expansion of Protein Structures and Interactions for Docking (PEPSI-Dock) improves the accuracy of the first stage of the docking pipeline, which will sharpen up the final predictions. Indeed, PEPSI-Dock benefits from the precision of a very detailed data-driven model of the binding free energy used with a global and exhaustive rigid-body search space. As well as being accurate, our computations are among the fastest by virtue of the sparse representation of the pre-computed potentials and FFT-accelerated sampling techniques. Overall, this is the first demonstration of a FFT-accelerated docking method coupled with an arbitrary-shaped distance-dependent interaction potential. RESULTS: First, we present a novel learning process to compute data-driven distant-dependent pairwise potentials, adapted from our previous method used for rescoring of putative protein-protein binding poses. The potential coefficients are learned by combining machine-learning techniques with physically interpretable descriptors. Then, we describe the integration of the deduced potentials into a FFT-accelerated spherical sampling provided by the Hex library. Overall, on a training set of 163 heterodimers, PEPSI-Dock achieves a success rate of 91% mid-quality predictions in the top-10 solutions. On a subset of the protein docking benchmark v5, it achieves 44.4% mid-quality predictions in the top-10 solutions when starting from bound structures and 20.5% when starting from unbound structures. The method runs in 5-15 min on a modern laptop and can easily be extended to other types of interactions. AVAILABILITY AND IMPLEMENTATION: https://team.inria.fr/nano-d/software/PEPSI-Dock CONTACT: [email protected].

MOTIVATION: Calculating multiple protein structure alignments (MSAs) is important for understanding functional and evolutionary relationships between protein families, and for modeling protein structures by homology. While incorporating backbone flexibility promises to circumvent many of the limitations of rigid MSA algorithms, very few flexible MSA algorithms exist today. This article describes several novel improvements to the Kpax algorithm which allow high quality flexible MSAs to be calculated. This article also introduces a new Gaussian-based MSA quality measure called 'M-score', which circumvents the pitfalls of RMSD-based quality measures. RESULTS: As well as calculating flexible MSAs, the new version of Kpax can also score MSAs from other aligners and from previously aligned reference datasets. Results are presented for a large-scale evaluation of the Homstrad, SABmark and SISY benchmark sets using Kpax and Matt as examples of state-of-the-art flexible aligners and 3DCOMB as an example of a state-of-the-art rigid aligner. These results demonstrate the utility of the M-score as a measure of MSA quality and show that high quality MSAs may be achieved when structural flexibility is properly taken into account. AVAILABILITY AND IMPLEMENTATION: Kpax 5.0 may be downloaded for academic use at http://kpax.loria.fr/ CONTACT: [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: Aligning and comparing protein structures is important for understanding their evolutionary and functional relationships. With the rapid growth of protein structure databases in recent years, the need to align, superpose and compare protein structures rapidly and accurately has never been greater. Many structural alignment algorithms have been described in the past 20 years. However, achieving an algorithm that is both accurate and fast remains a considerable challenge. RESULTS: We have developed a novel protein structure alignment algorithm called 'Kpax', which exploits the highly predictable covalent geometry of C(α) atoms to define multiple local coordinate frames in which backbone peptide fragments may be oriented and compared using sensitive Gaussian overlap scoring functions. A global alignment and hence a structural superposition may then be found rapidly using dynamic programming with secondary structure-specific gap penalties. When superposing pairs of structures, Kpax tends to give tighter secondary structure overlays than several popular structure alignment algorithms. When searching the CATH database, Kpax is faster and more accurate than the very efficient Yakusa algorithm, and it gives almost the same high level of fold recognition as TM-Align while being more than 100 times faster.

MOTIVATION: In recent years, much structural information on protein domains and their pair-wise interactions has been made available in public databases. However, it is not yet clear how best to use this information to discover general rules or interaction patterns about structural protein-protein interactions. Improving our ability to detect and exploit structural interaction patterns will help to provide a better 3D picture of the known protein interactome, and will help to guide docking-based predictions of the 3D structures of unsolved protein complexes. RESULTS: This article presents KBDOCK, a 3D database approach for spatially clustering protein binding sites and for performing template-based (knowledge-based) protein docking. KBDOCK combines residue contact information from the 3DID database with the Pfam protein domain family classification together with coordinate data from the Protein Data Bank. This allows the 3D configurations of all known hetero domain-domain interactions to be superposed and clustered for each Pfam family. We find that most Pfam domain families have up to four hetero binding sites, and over 60% of all domain families have just one hetero binding site. The utility of this approach for template-based docking is demonstrated using 73 complexes from the Protein Docking Benchmark. Overall, up to 45 out of 73 complexes may be modelled by direct homology to existing domain interfaces, and key binding site information is found for 24 of the 28 remaining complexes. These results show that KBDOCK can often provide useful information for predicting the structures of unknown protein complexes. AVAILABILITY: http://kbdock.loria.fr/ CONTACT: [email protected] SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

MOTIVATION: Modelling protein-protein interactions (PPIs) is an increasingly important aspect of structural bioinformatics. However, predicting PPIs using in silico docking techniques is computationally very expensive. Developing very fast protein docking tools will be useful for studying large-scale PPI networks, and could contribute to the rational design of new drugs. RESULTS: The Hex spherical polar Fourier protein docking algorithm has been implemented on Nvidia graphics processor units (GPUs). On a GTX 285 GPU, an exhaustive and densely sampled 6D docking search can be calculated in just 15 s using multiple 1D fast Fourier transforms (FFTs). This represents a 45-fold speed-up over the corresponding calculation on a single CPU, being at least two orders of magnitude times faster than a similar CPU calculation using ZDOCK 3.0.1, and estimated to be at least three orders of magnitude faster than the GPU-accelerated version of PIPER on comparable hardware. Hence, for the first time, exhaustive FFT-based protein docking calculations may now be performed in a matter of seconds on a contemporary GPU. Three-dimensional Hex FFT correlations are also accelerated by the GPU, but the speed-up factor of only 2.5 is much less than that obtained with 1D FFTs. Thus, the Hex algorithm appears to be especially well suited to exploit GPUs compared to conventional 3D FFT docking approaches. AVAILABILITY: http://hex.loria.fr/ and http://hexserver.loria.fr/.

MOTIVATION: Predicting how proteins interact at the molecular level is a computationally intensive task. Many protein docking algorithms begin by using fast Fourier transform (FFT) correlation techniques to find putative rigid body docking orientations. Most such approaches use 3D Cartesian grids and are therefore limited to computing three dimensional (3D) translational correlations. However, translational FFTs can speed up the calculation in only three of the six rigid body degrees of freedom, and they cannot easily incorporate prior knowledge about a complex to focus and hence further accelerate the calculation. Furthemore, several groups have developed multi-term interaction potentials and others use multi-copy approaches to simulate protein flexibility, which both add to the computational cost of FFT-based docking algorithms. Hence there is a need to develop more powerful and more versatile FFT docking techniques. RESULTS: This article presents a closed-form 6D spherical polar Fourier correlation expression from which arbitrary multi-dimensional multi-property multi-resolution FFT correlations may be generated. The approach is demonstrated by calculating 1D, 3D and 5D rotational correlations of 3D shape and electrostatic expansions up to polynomial order L=30 on a 2 GB personal computer. As expected, 3D correlations are found to be considerably faster than 1D correlations but, surprisingly, 5D correlations are often slower than 3D correlations. Nonetheless, we show that 5D correlations will be advantageous when calculating multi-term knowledge-based interaction potentials. When docking the 84 complexes of the Protein Docking Benchmark, blind 3D shape plus electrostatic correlations take around 30 minutes on a contemporary personal computer and find acceptable solutions within the top 20 in 16 cases. Applying a simple angular constraint to focus the calculation around the receptor binding site produces acceptable solutions within the top 20 in 28 cases. Further constraining the search to the ligand binding site gives up to 48 solutions within the top 20, with calculation times of just a few minutes per complex. Hence the approach described provides a practical and fast tool for rigid body protein-protein docking, especially when prior knowledge about one or both binding sites is available.