Michal Brylinski

A threading-based method (FINDSITE) for ligand-binding site prediction and functional annotation.

The detection of ligand-binding sites is often the starting point for protein function identification and drug discovery. Because of inaccuracies in predicted protein structures, extant binding pocket-detection methods are limited to experimentally solved structures. Here, FINDSITE, a method for ligand-binding site prediction and functional annotation based on binding-site similarity across groups of weakly homologous template structures identified from threading, is described. For crystal structures, considering a cutoff distance of 4 Å as the hit criterion, the success rate is 70.9% for identifying the best of top five predicted ligand-binding sites with a ranking accuracy of 76.0%. Both high prediction accuracy and ability to correctly rank identified binding sites are sustained when approximate protein models (<35% sequence identity to the closest template structure) are used, showing a 67.3% success rate with 75.5% ranking accuracy. In practice, FINDSITE tolerates structural inaccuracies in protein models up to a rmsd from the crystal structure of 8–10 Å. This is because analysis of weakly homologous protein models reveals that about half have a rmsd from the native binding site <2 Å. Furthermore, the chemical properties of template-bound ligands can be used to select ligand templates associated with the binding site. In most cases, FINDSITE can accurately assign a molecular function to the protein model.

FINDSITE: a combined evolution/structure-based approach to protein function prediction

A key challenge of the post-genomic era is the identification of the function(s) of all the molecules in a given organism. Here, we review the status of sequence and structure-based approaches to protein function inference and ligand screening that can provide functional insights for a significant fraction of the approximately 50% of ORFs of unassigned function in an average proteome. We then describe FINDSITE, a recently developed algorithm for ligand binding site prediction, ligand screening and molecular function prediction, which is based on binding site conservation across evolutionary distant proteins identified by threading. Importantly, FINDSITE gives comparable results when high-resolution experimental structures as well as predicted protein models are used.

The continuity of protein structure space is an intrinsic property of proteins

The classical view of the space of protein structures is that it is populated by a discrete set of protein folds. For proteins up to 200 residues long, by using structural alignments and building upon ideas of the completeness and continuity of structure space, we show that nearly any structure is significantly related to any other using a transitive set of no more than 7 intermediate structurally related proteins. This result holds for all structures in the Protein Data Bank, even when structural relationships between evolutionary related proteins (as detected by threading or functional analyses) are excluded. A similar picture holds for an artificial library of compact, hydrogen-bonded, homopolypeptide structures. The 3 sets share the global connectivity features of random graphs, in which the local connectivity of each node (i.e., the number of neighboring structures per protein) is preserved. This high connectivity supports the continuous view of single-domain protein structure space. More importantly, these results do not depend on evolution, rather just on the physics of protein structures. The fact that evolutionary divergence need not be invoked to explain the continuous nature of protein structure space has implications for how the universe of protein structures might have originated, and how function should be transferred between proteins of similar structure.

Comprehensive structural and functional characterization of the human kinome by protein structure modeling and ligand virtual screening.

The growing interest in the identification of kinase inhibitors, promising therapeutics in the treatment of many diseases, has created a demand for the structural characterization of the entire human kinome. At the outset of the drug development process, the lead-finding stage, approaches that enrich the screening library with bioactive compounds are needed. Here, protein structure based methods can play an important role, but despite structural genomics efforts, it is unlikely that the three-dimensional structures of the entire kinome will be available soon. Therefore, at the proteome level, structure-based approaches must rely on predicted models, with a key issue being their utility in virtual ligand screening. In this study, we employ the recently developed FINDSITE/Q-Dock ligand homology modeling approach, which is well-suited for proteome-scale applications using predicted structures, to provide extensive structural and functional characterization of the human kinome. Specifically, we construct structure models for the human kinome; these are subsequently subject to virtual screening against a library of more than 2 million compounds. To rank the compounds, we employ a hierarchical approach that combines ligand- and structure-based filters. Modeling accuracy is carefully validated using available experimental data with particularly encouraging results found for the ability to identify, without prior knowledge, specific kinase inhibitors. More generally, the modeling procedure results in a large number of predicted molecular interactions between kinases and small ligands that should be of practical use in the development of novel inhibitors. The data set is freely available to the academic community via a user-friendly Web interface at http://cssb.biology.gatech.edu/kinomelhm/ as well as at the ZINC Web site ( http://zinc.docking.org/applications/2010Apr/Brylinski-2010.tar.gz ).

eThread: A Highly Optimized Machine Learning-Based Approach to Meta-Threading and the Modeling of Protein Tertiary Structures

Template-based modeling that employs various meta-threading techniques is currently the most accurate, and consequently the most commonly used, approach for protein structure prediction. Despite the evident progress in this field, accurate structure models cannot be constructed for a significant fraction of gene products, thus the development of new algorithms is required. Here, we describe the development, optimization and large-scale benchmarking of eThread, a highly accurate meta-threading procedure for the identification of structural templates and the construction of corresponding target-to-template alignments. eThread integrates ten state-of-the-art threading/fold recognition algorithms in a local environment and extensively uses various machine learning techniques to carry out fully automated template-based protein structure modeling. Tertiary structure prediction employs two protocols based on widely used modeling algorithms: Modeller and TASSER-Lite. As a part of eThread, we also developed eContact, which is a Bayesian classifier for the prediction of inter-residue contacts and eRank, which effectively ranks generated multiple protein models and provides reliable confidence estimates as structure quality assessment. Excluding closely related templates from the modeling process, eThread generates models, which are correct at the fold level, for >80% of the targets; 40–50% of the constructed models are of a very high quality, which would be considered accurate at the family level. Furthermore, in large-scale benchmarking, we compare the performance of eThread to several alternative methods commonly used in protein structure prediction. Finally, we estimate the upper bound for this type of approach and discuss the directions towards further improvements.

Further evidence for the likely completeness of the library of solved single domain protein structures.

Recent studies questioned whether the Protein Data Bank (PDB) contains all compact, single domain protein structures. Here, we show that all quasi-spherical, QS, random protein structures devoid of secondary structure are in the PDB and are excellent templates for all native PDB proteins up to 250 residues. Because QS templates have a similar global contour as native, TASSER can refine 98% (90%) of those whose TM-score is 0.4 (0.35) to structures greater than or equal to the 0.5 TM-score threshold (0.74 (0.64) mean TM-score) for CATH/SCOP assignment. On the basis of this and the fact that, at a TM-score of 0.4, 83% (90%) of all (internal) core secondary structure elements are recovered, a 0.40 TM-score is an appropriate fold similarity assignment threshold. Despite the claims of Taylor, Trovato, and Zhou that many of their structures lack a PDB counterpart, using fr-TM-align, at a 0.45 (0.5) TM-score threshold, essentially all (most) are found in the PDB. Thus, the conclusion that the PDB is likely complete is further supported.

Nonlinear scoring functions for similarity-based ligand docking and binding affinity prediction.

A common strategy for virtual screening considers a systematic docking of a large library of organic compounds into the target sites in protein receptors with promising leads selected based on favorable intermolecular interactions. Despite a continuous progress in the modeling of protein-ligand interactions for pharmaceutical design, important challenges still remain, thus the development of novel techniques is required. In this communication, we describe eSimDock, a new approach to ligand docking and binding affinity prediction. eSimDock employs nonlinear machine learning-based scoring functions to improve the accuracy of ligand ranking and similarity-based binding pose prediction, and to increase the tolerance to structural imperfections in the target structures. In large-scale benchmarking using the Astex/CCDC data set, we show that 53.9% (67.9%) of the predicted ligand poses have RMSD of <2 Å (<3 Å). Moreover, using binding sites predicted by recently developed eFindSite, eSimDock models ligand binding poses with an RMSD of 4 Å for 50.0-39.7% of the complexes at the protein homology level limited to 80-40%. Simulations against non-native receptor structures, whose mean backbone rearrangements vary from 0.5 to 5.0 Å Cα-RMSD, show that the ratio of docking accuracy and the estimated upper bound is at a constant level of ∼0.65. Pearson correlation coefficient between experimental and predicted by eSimDock Ki values for a large data set of the crystal structures of protein-ligand complexes from BindingDB is 0.58, which decreases only to 0.46 when target structures distorted to 3.0 Å Cα-RMSD are used. Finally, two case studies demonstrate that eSimDock can be customized to specific applications as well. These encouraging results show that the performance of eSimDock is largely unaffected by the deformations of ligand binding regions, thus it represents a practical strategy for across-proteome virtual screening using protein models. eSimDock is freely available to the academic community as a Web server at http://www.brylinski.org/esimdock .

Use of protein cross-linking and radiolytic footprinting to elucidate PsbP and PsbQ interactions within higher plant Photosystem II

Significance In higher plant Photosystem II, the PsbP and PsbQ proteins provide critical support for oxygen evolution at physiological calcium and chloride concentrations. The locations of these components within the photosystem, however, are unclear. Our findings that (i) the N terminus of PsbP, which is unresolved in the current high-resolution structure of this subunit, forms a compact structure and associates with the C-terminal domain of the protein and (ii) PsbP and PsbQ directly interact to form a framework for understanding the organization of these subunits within the higher plant photosystem. Protein cross-linking and radiolytic footprinting coupled with high-resolution mass spectrometry were used to examine the structure of PsbP and PsbQ when they are bound to Photosystem II. In its bound state, the N-terminal 15-amino-acid residue domain of PsbP, which is unresolved in current crystal structures, interacts with domains in the C terminus of the protein. These interactions may serve to stabilize the structure of the N terminus and may facilitate PsbP binding and function. These interactions place strong structural constraints on the organization of PsbP when associated with the Photosystem II complex. Additionally, amino acid residues in the structurally unresolved loop 3A domain of PsbP (90K–107V), 93Y and 96K, are in close proximity (≤11.4 Å) to the N-terminal 1E residue of PsbQ. These findings are the first, to our knowledge, to identify a putative region of interaction between these two components. Cross-linked domains within PsbQ were also identified, indicating that two PsbQ molecules can interact in higher plants in a manner similar to that observed by Liu et al. [(2014) Proc Natl Acad Sci 111(12):4638–4643] in cyanobacterial Photosystem II. This interaction is consistent with either intra-Photosystem II dimer or inter-Photosystem II dimer models in higher plants. Finally, OH• produced by synchrotron radiolysis of water was used to oxidatively modify surface residues on PsbP and PsbQ. Domains on the surface of both protein subunits were resistant to modification, indicating that they were shielded from water and appear to define buried regions that are in contact with other Photosystem II components.

Early-Stage Folding in Proteins (In Silico) Sequence-to-Structure Relation

A sequence-to-structure library has been created based on the complete PDB database. The tetrapeptide was selected as a unit representing a well-defined structural motif. Seven structural forms were introduced for structure classification. The early-stage folding conformations were used as the objects for structure analysis and classification. The degree of determinability was estimated for the sequence-to-structure and structure-to-sequence relations. Probability calculus and informational entropy were applied for quantitative estimation of the mutual relation between them. The structural motifs representing different forms of loops and bends were found to favor particular sequences in structure-to-sequence analysis.

Predicting protein interface residues using easily accessible on-line resources

It has been more than a decade since the completion of the Human Genome Project that provided us with a complete list of human proteins. The next obvious task is to figure out how various parts interact with each other. On that account, we review 10 methods for protein interface prediction, which are freely available as web servers. In addition, we comparatively evaluate their performance on a common data set comprising different quality target structures. We find that using experimental structures and high-quality homology models, structure-based methods outperform those using only protein sequences, with global template-based approaches providing the best performance. For moderate-quality models, sequence-based methods often perform better than those structure-based techniques that rely on fine atomic details. We note that post-processing protocols implemented in several methods quantitatively improve the results only for experimental structures, suggesting that these procedures should be tuned up for computer-generated models. Finally, we anticipate that advanced meta-prediction protocols are likely to enhance interface residue prediction. Notwithstanding further improvements, easily accessible web servers already provide the scientific community with convenient resources for the identification of protein-protein interaction sites.