Knowledge-based Scoring Function Research Articles

New Knowledge-Based Scoring Function with Inclusion of Backbone Conformational Entropies from Protein Structures.

Accurate prediction of a protein's structure requires a reliable free energy function that consists of both enthalpic and entropic contributions. Although considerable progresses have been made in the calculation of potential energies in protein structure prediction, the computation for entropies of protein has lagged far behind, due to the challenge that estimation of entropies often requires expensive conformational sampling. In this study, we have used a knowledge-based approach to estimate the backbone conformational entropies from experimentally determined structures. Instead of conducting computationally expensive MD/MC simulations, we obtained the entropies of protein structures based on the normalized probability distributions of back dihedral angles observed in the native structures. Our new knowledge-based scoring function with inclusion of the backbone entropies, which is referred to as ITScoreDA or ITDA, was extensively evaluated on 16 commonly used decoy sets and compared with 50 other published scoring functions. It was shown that ITDA is significantly superior to the other tested scoring functions in selecting native structures from decoys. The present study suggests the role of backbone conformational entropies in protein structures and provides a way for fast estimation of the entropic effect.

Lessons learned from participating in D3R 2016 Grand Challenge 2: compounds targeting the farnesoid X receptor.

D3R 2016 Grand Challenge 2 focused on predictions of binding modes and affinities for 102 compounds against the farnesoid X receptor (FXR). In this challenge, two distinct methods, a docking-based method and a template-based method, were employed by our team for the binding mode prediction. For the new template-based method, 3D ligand similarities were calculated for each query compound against the ligands in the co-crystal structures of FXR available in Protein Data Bank. The binding mode was predicted based on the co-crystal protein structure containing the ligand with the best ligand similarity score against the query compound. For the FXR dataset, the template-based method achieved a better performance than the docking-based method on the binding mode prediction. For the binding affinity prediction, an in-house knowledge-based scoring function ITScore2 and MM/PBSA approach were employed. Good performance was achieved for MM/PBSA, whereas the performance of ITScore2 was sensitive to ligand composition, e.g. the percentage of carbon atoms in the compounds. The sensitivity to ligand composition could be a clue for the further improvement of our knowledge-based scoring function.

Exploring potent ligand for proteins: insights from knowledge-based scoring functions and molecular interaction energies

Two different scoring functions, Hirshfeld fingerprint-based scoring (HFBS) and molecular operating environment (MOE), and the kernel energy method (KEM) along with counterpoise (CP)-corrected approach were used to estimate the binding energies of protein–ligand complexes and tested against a series of inhibitors of human aldose reductase enzyme. The new scoring function, HFBS, is based on Hirshfeld fingerprints, which are 2D histogram plots of the distances from the molecular Hirshfeld surface to the nearest atomic nuclei inside versus outside the surface and are highly sensitive to the immediate environment of the molecule. The Hirshfeld surface plotted over the ligand molecule helped to visualize the contacts with the active site residues and solvents, which were then taken into account for interaction energy calculations. Application of KEM-assisted CP-corrected approach facilitated an efficient way of calculating interaction energies in protein complex systems. Interaction energies calculated using MP2/6-31G(d) level of theory allowed us to rank the ligands by potency. We find that both the KEM-assisted CP-corrected interaction energies and the scoring functions used here predict comparable rankings for the strength of binding of the series of ligands as docked to the active site of the protein, which are also in good agreement with the experimental binding affinities in this case.

SPA-LN: a scoring function of ligand-nucleic acid interactions via optimizing both specificity and affinity.

Nucleic acids have been widely recognized as potential targets in drug discovery and aptamer selection. Quantifying the interactions between small molecules and nucleic acids is critical to discover lead compounds and design novel aptamers. Scoring function is normally employed to quantify the interactions in structure-based virtual screening. However, the predictive power of nucleic acid–ligand scoring functions is still a challenge compared to other types of biomolecular recognition. With the rapid growth of experimentally determined nucleic acid–ligand complex structures, in this work, we develop a knowledge-based scoring function of nucleic acid–ligand interactions, namely SPA-LN. SPA-LN is optimized by maximizing both the affinity and specificity of native complex structures. The development strategy is different from those of previous nucleic acid–ligand scoring functions which focus on the affinity only in the optimization. The native conformation is stabilized while non-native conformations are destabilized by our optimization, making the funnel-like binding energy landscape more biased toward the native state. The performance of SPA-LN validates the development strategy and provides a relatively more accurate way to score the nucleic acid–ligand interactions.

A Hybrid Knowledge-Based and Empirical Scoring Function for Protein–Ligand Interaction: SMoG2016

We present the third generation of our scoring function for the prediction of protein-ligand binding free energy. This function is now a hybrid between a knowledge-based potential and an empirical function. We constructed a diversified set of ∼1000 complexes from the PDBBinding-CN database for the training of the function, and we show that this number of complexes generates enough data to build the potential. The occurrence of 420 different types of atomic pairwise interactions is computed in up to five different ranges of distances to derive the knowledge-based part. All of the parameters were optimized, and we were able to considerably improve the accuracy of the scoring function with a Pearson correlation coefficient against experimental binding free energies of up to 0.57, which ranks our new scoring function as one of the best currently available and the second-best in terms of standard deviation (SD = 1.68 kcal/mol). The function was then further improved by inclusion of different terms taking into account repulsion and loss of entropy upon binding, and we show that it is capable of recovering native binding poses up to 80% of the time. All of the programs, tools, and protein sets are released in the Supporting Information or as open-source programs.

Structural and energy determinants in protein-RNA docking.

Deciphering the structural and energetic determinants of protein-RNA interactions harbors the potential to understand key cell processes at molecular level, such as gene expression and regulation. With this purpose, computational methods like docking aim to complement current biophysical and structural biology efforts. However, the few reported docking algorithms for protein-RNA interactions show limited predictive success rates, mainly due to incomplete sampling of the conformational space of both the protein and the RNA molecules, as well as to the difficulties of the scoring function in identifying the correct docking models. Here, we have tested the predictive value of a variety of knowledge-based and energetic scoring functions on a recently published protein-RNA docking benchmark and developed a scoring function able to efficiently discriminate docking decoys. We first performed docking calculations with the bound conformation, which allowed us to analyze the problem in optimal conditions. We found that geometry-based terms and electrostatics were the most important scoring terms, while binding propensities and desolvation were much less relevant for the scoring of protein-RNA models. This is in contrast with what we observed for protein-protein docking. The results also showed an interesting dependence of the predictive rates on the flexibility of the protein molecule, which arises from the observed higher positive charge of flexible interfaces and provides hints for future development of more efficient protein-RNA docking methods.

Human and server docking prediction for CAPRI round 30-35 using LZerD with combined scoring functions.

We report the performance of protein-protein docking predictions by our group for recent rounds of the Critical Assessment of Prediction of Interactions (CAPRI), a community-wide assessment of state-of-the-art docking methods. Our prediction procedure uses a protein-protein docking program named LZerD developed in our group. LZerD represents a protein surface with 3D Zernike descriptors (3DZD), which are based on a mathematical series expansion of a 3D function. The appropriate soft representation of protein surface with 3DZD makes the method more tolerant to conformational change of proteins upon docking, which adds an advantage for unbound docking. Docking was guided by interface residue prediction performed with BindML and cons-PPISP as well as literature information when available. The generated docking models were ranked by a combination of scoring functions, including PRESCO, which evaluates the native-likeness of residues' spatial environments in structure models. First, we discuss the overall performance of our group in the CAPRI prediction rounds and investigate the reasons for unsuccessful cases. Then, we examine the performance of several knowledge-based scoring functions and their combinations for ranking docking models. It was found that the quality of a pool of docking models generated by LZerD, that is whether or not the pool includes near-native models, can be predicted by the correlation of multiple scores. Although the current analysis used docking models generated by LZerD, findings on scoring functions are expected to be universally applicable to other docking methods. Proteins 2017; 85:513-527. © 2016 Wiley Periodicals, Inc.

Docking pose selection by interaction pattern graph similarity: application to the D3R grand challenge 2015.

High affinity ligands for a given target tend to share key molecular interactions with important anchoring amino acids and therefore often present quite conserved interaction patterns. This simple concept was formalized in a topological knowledge-based scoring function (GRIM) for selecting the most appropriate docking poses from previously X-rayed interaction patterns. GRIM first converts protein-ligand atomic coordinates (docking poses) into a simple 3D graph describing the corresponding interaction pattern. In a second step, proposed graphs are compared to that found from template structures in the Protein Data Bank. Last, all docking poses are rescored according to an empirical score (GRIMscore) accounting for overlap of maximum common subgraphs. Taking the opportunity of the public D3R Grand Challenge 2015, GRIM was used to rescore docking poses for 36 ligands (6 HSP90α inhibitors, 30 MAP4K4 inhibitors) prior to the release of the corresponding protein-ligand X-ray structures. When applied to the HSP90α dataset, for which many protein-ligand X-ray structures are already available, GRIM provided very high quality solutions (mean rmsd=1.06Å, n=6) as top-ranked poses, and significantly outperformed a state-of-the-art scoring function. In the case of MAP4K4 inhibitors, for which preexisting 3D knowledge is scarce and chemical diversity is much larger, the accuracy of GRIM poses decays (mean rmsd=3.18 Å, n=30) although GRIM still outperforms an energy-based scoring function. GRIM rescoring appears to be quite robust with comparison to the other approaches competing for the same challenge (42 submissions for the HSP90 dataset, 27 for the MAP4K4 dataset) as it ranked 3rd and 2nd respectively, for the two investigated datasets. The rescoring method is quite simple to implement, independent on a docking engine, and applicable to any target for which at least one holo X-ray structure is available.

A robust algorithm for optimizing protein structures with NMR chemical shifts.

Over the past decade, a number of methods have been developed to determine the approximate structure of proteins using minimal NMR experimental information such as chemical shifts alone, sparse NOEs alone or a combination of comparative modeling data and chemical shifts. However, there have been relatively few methods that allow these approximate models to be substantively refined or improved using the available NMR chemical shift data. Here, we present a novel method, called Chemical Shift driven Genetic Algorithm for biased Molecular Dynamics (CS-GAMDy), for the robust optimization of protein structures using experimental NMR chemical shifts. The method incorporates knowledge-based scoring functions and structural information derived from NMR chemical shifts via a unique combination of multi-objective MD biasing, a genetic algorithm, and the widely used XPLOR molecular modelling language. Using this approach, we demonstrate that CS-GAMDy is able to refine and/or fold models that are as much as 10 Å (RMSD) away from the correct structure using only NMR chemical shift data. CS-GAMDy is also able to refine of a wide range of approximate or mildly erroneous protein structures to more closely match the known/correct structure and the known/correct chemical shifts. We believe CS-GAMDy will allow protein models generated by sparse restraint or chemical-shift-only methods to achieve sufficiently high quality to be considered fully refined and "PDB worthy". The CS-GAMDy algorithm is explained in detail and its performance is compared over a range of refinement scenarios with several commonly used protein structure refinement protocols. The program has been designed to be easily installed and easily used and is available at http://www.gamdy.ca.

An accurate metalloprotein-specific scoring function and molecular docking program devised by a dynamic sampling and iteration optimization strategy.

Metalloproteins, particularly zinc metalloproteins, are promising therapeutic targets, and recent efforts have focused on the identification of potent and selective inhibitors of these proteins. However, the ability of current drug discovery and design technologies, such as molecular docking and molecular dynamics simulations, to probe metal-ligand interactions remains limited because of their complicated coordination geometries and rough treatment in current force fields. Herein we introduce a robust, multiobjective optimization algorithm-driven metalloprotein-specific docking program named MpSDock, which runs on a scheme similar to consensus scoring consisting of a force-field-based scoring function and a knowledge-based scoring function. For this purpose, in this study, an effective knowledge-based zinc metalloprotein-specific scoring function based on the inverse Boltzmann law was designed and optimized using a dynamic sampling and iteration optimization strategy. This optimization strategy can dynamically sample and regenerate decoy poses used in each iteration step of refining the scoring function, thus dramatically improving both the effectiveness of the exploration of the binding conformational space and the sensitivity of the ranking of the native binding poses. To validate the zinc metalloprotein-specific scoring function and its special built-in docking program, denoted MpSDockZn, an extensive comparison was performed against six universal, popular docking programs: Glide XP mode, Glide SP mode, Gold, AutoDock, AutoDock4Zn, and EADock DSS. The zinc metalloprotein-specific knowledge-based scoring function exhibited prominent performance in accurately describing the geometries and interactions of the coordination bonds between the zinc ions and chelating agents of the ligands. In addition, MpSDockZn had a competitive ability to sample and identify native binding poses with a higher success rate than the other six docking programs.

Computational modeling of membrane proteins.

The determination of membrane protein (MP) structures has always trailed that of soluble proteins due to difficulties in their overexpression, reconstitution into membrane mimetics, and subsequent structure determination. The percentage of MP structures in the protein databank (PDB) has been at a constant 1-2% for the last decade. In contrast, over half of all drugs target MPs, only highlighting how little we understand about drug-specific effects in the human body. To reduce this gap, researchers have attempted to predict structural features of MPs even before the first structure was experimentally elucidated. In this review, we present current computational methods to predict MP structure, starting with secondary structure prediction, prediction of trans-membrane spans, and topology. Even though these methods generate reliable predictions, challenges such as predicting kinks or precise beginnings and ends of secondary structure elements are still waiting to be addressed. We describe recent developments in the prediction of 3D structures of both α-helical MPs as well as β-barrels using comparative modeling techniques, de novo methods, and molecular dynamics (MD) simulations. The increase of MP structures has (1) facilitated comparative modeling due to availability of more and better templates, and (2) improved the statistics for knowledge-based scoring functions. Moreover, de novo methods have benefited from the use of correlated mutations as restraints. Finally, we outline current advances that will likely shape the field in the forthcoming decade.

The role of nucleobase interactions in RNA structure and dynamics.

The intricate network of interactions observed in RNA three-dimensional structures is often described in terms of a multitude of geometrical properties, including helical parameters, base pairing/stacking, hydrogen bonding and backbone conformation. We show that a simple molecular representation consisting in one oriented bead per nucleotide can account for the fundamental structural properties of RNA. In this framework, canonical Watson-Crick, non-Watson-Crick base-pairing and base-stacking interactions can be unambiguously identified within a well-defined interaction shell. We validate this representation by performing two independent, complementary tests. First, we use it to construct a sequence-independent, knowledge-based scoring function for RNA structural prediction, which compares favorably to fully atomistic, state-of-the-art techniques. Second, we define a metric to measure deviation between RNA structures that directly reports on the differences in the base–base interaction network. The effectiveness of this metric is tested with respect to the ability to discriminate between structurally and kinetically distant RNA conformations, performing better compared to standard techniques. Taken together, our results suggest that this minimalist, nucleobase-centric representation captures the main interactions that are relevant for describing RNA structure and dynamics.

Structure-Based Multiscale Approach for Identification of Interaction Partners of PDZ Domains

PDZ domains are peptide recognition modules which mediate specific protein-protein interactions and are known to have a complex specificity landscape. We have developed a novel structure-based multiscale approach which identifies crucial specificity determining residues (SDRs) of PDZ domains from explicit solvent molecular dynamics (MD) simulations on PDZ-peptide complexes and uses these SDRs in combination with knowledge-based scoring functions for proteomewide identification of their interaction partners. Multiple explicit solvent simulations ranging from 5 to 50 ns duration have been carried out on 28 PDZ-peptide complexes with known binding affinities. MM/PBSA binding energy values calculated from these simulations show a correlation coefficient of 0.755 with the experimental binding affinities. On the basis of the SDRs of PDZ domains identified by MD simulations, we have developed a simple scoring scheme for evaluating binding energies for PDZ-peptide complexes using residue based statistical pair potentials. This multiscale approach has been benchmarked on a mouse PDZ proteome array data set by calculating the binding energies for 217 different substrate peptides in binding pockets of 64 different mouse PDZ domains. Receiver operating characteristic (ROC) curve analysis indicates that, the area under curve (AUC) values for binder vs nonbinder classification by our structure based method is 0.780. Our structure based method does not require experimental PDZ-peptide binding data for training.

A Bayesian statistical approach of improving knowledge‐based scoring functions for protein–ligand interactions

Knowledge-based scoring functions are widely used for assessing putative complexes in protein-ligand and protein-protein docking and for structure prediction. Even with large training sets, knowledge-based scoring functions face the inevitable problem of sparse data. Here, we have developed a novel approach for handling the sparse data problem that is based on estimating the inaccuracies in knowledge-based scoring functions. This inaccuracy estimation is used to automatically weight the knowledge-based scoring function with an alternative, force-field-based potential (FFP) that does not rely on training data and can, therefore, provide an improved approximation of the interactions between rare chemical groups. The current version of STScore, a protein-ligand scoring function using our method, achieves a binding mode prediction success rate of 91% on the set of 100 complexes by Wang et al., and a binding affinity correlation of 0.514 with the experimentally determined affinities in PDBbind. The method presented here may be used with other FFPs and other knowledge-based scoring functions and can also be applied to protein-protein docking and protein structure prediction.

A knowledge-based scoring function for protein-RNA interactions derived from a statistical mechanics-based iterative method

Protein-RNA interactions play important roles in many biological processes. Given the high cost and technique difficulties in experimental methods, computationally predicting the binding complexes from individual protein and RNA structures is pressingly needed, in which a reliable scoring function is one of the critical components. Here, we have developed a knowledge-based scoring function, referred to as ITScore-PR, for protein-RNA binding mode prediction by using a statistical mechanics-based iterative method. The pairwise distance-dependent atomic interaction potentials of ITScore-PR were derived from experimentally determined protein–RNA complex structures. For validation, we have compared ITScore-PR with 10 other scoring methods on four diverse test sets. For bound docking, ITScore-PR achieved a success rate of up to 86% if the top prediction was considered and up to 94% if the top 10 predictions were considered, respectively. For truly unbound docking, the respective success rates of ITScore-PR were up to 24 and 46%. ITScore-PR can be used stand-alone or easily implemented in other docking programs for protein–RNA recognition.

A Global Machine Learning Based Scoring Function for Protein Structure Prediction

We present a knowledge-based function to score protein decoys based on their similarity to native structure. A set of features is constructed to describe the structure and sequence of the entire protein chain. Furthermore, a qualitative relationship is established between the calculated features and the underlying electromagnetic interaction that dominates this scale. The features we use are associated with residue-residue distances, residue-solvent distances, pairwise knowledge-based potentials and a four-body potential. In addition, we introduce a new target to be predicted, the fitness score, which measures the similarity of a model to the native structure. This new approach enables us to obtain information both from decoys and from native structures. It is also devoid of previous problems associated with knowledge-based potentials. These features were obtained for a large set of native and decoy structures and a back-propagating neural network was trained to predict the fitness score. Overall this new scoring potential proved to be superior to the knowledge-based scoring functions used as its inputs. In particular, in the latest CASP (CASP10) experiment our method was ranked third for all targets, and second for freely modeled hard targets among about 200 groups for top model prediction. Ours was the only method ranked in the top three for all targets and for hard targets. This shows that initial results from the novel approach are able to capture details that were missed by a broad spectrum of protein structure prediction approaches. Source codes and executable from this work are freely available at http://mathmed.org/#Software and http://mamiris.com/.

PiDNA: predicting protein–DNA interactions with structural models

Predicting binding sites of a transcription factor in the genome is an important, but challenging, issue in studying gene regulation. In the past decade, a large number of protein–DNA co-crystallized structures available in the Protein Data Bank have facilitated the understanding of interacting mechanisms between transcription factors and their binding sites. Recent studies have shown that both physics-based and knowledge-based potential functions can be applied to protein–DNA complex structures to deliver position weight matrices (PWMs) that are consistent with the experimental data. To further use the available structural models, the proposed Web server, PiDNA, aims at first constructing reliable PWMs by applying an atomic-level knowledge-based scoring function on numerous in silico mutated complex structures, and then using the PWM constructed by the structure models with small energy changes to predict the interaction between proteins and DNA sequences. With PiDNA, the users can easily predict the relative preference of all the DNA sequences with limited mutations from the native sequence co-crystallized in the model in a single run. More predictions on sequences with unlimited mutations can be realized by additional requests or file uploading. Three types of information can be downloaded after prediction: (i) the ranked list of mutated sequences, (ii) the PWM constructed by the favourable mutated structures, and (iii) any mutated protein–DNA complex structure models specified by the user. This study first shows that the constructed PWMs are similar to the annotated PWMs collected from databases or literature. Second, the prediction accuracy of PiDNA in detecting relatively high-specificity sites is evaluated by comparing the ranked lists against in vitro experiments from protein-binding microarrays. Finally, PiDNA is shown to be able to select the experimentally validated binding sites from 10 000 random sites with high accuracy. With PiDNA, the users can design biological experiments based on the predicted sequence specificity and/or request mutated structure models for further protein design. As well, it is expected that PiDNA can be incorporated with chromatin immunoprecipitation data to refine large-scale inference of in vivo protein–DNA interactions. PiDNA is available at: http://dna.bime.ntu.edu.tw/pidna.

ParaDockS - an open-source framework for molecular docking: implementation of target-class-specific scoring methods

Accurate scoring of protein-ligand interactions in molecular docking and virtual screening is still challenging. Despite great efforts, the performance of existing scoring functions strongly depends on the target structure under investigation. Recent developments in the direction of target-class-specific scoring methods and machine-learning-based classification models reveals a significant improvement in binding mode and activity prediction [1]. However, there is currently no open-source framework available that combines novel scoring techniques with molecular docking algorithms to make this simply applicable for virtual screening. Therefore, the aim of this work is the implementation and validation of target-biased scoring methods within the open-source docking framework ParaDockS [2]. ParaDockS includes algorithms for protein-ligand docking and is organized that every newly developed scoring function can be immediately implemented. Furthermore, interaction-based classifier, trained on a target-specific knowledge base can be used in a post-docking filter step. In general, we focus on knowledge-based scoring functions based on well established statistical potentials. Recently it was shown, that atom-pair potentials are also useful for training machine-learning models based on support vector machines or random forest approaches [1]. Such methods circumvent a particular functional form for the scoring function and thereby implicitly capture binding contributions that are hard to model explicitly. In a first validation study, we applied newly developed target-specific potentials on different kinase data sets and we found that they outperform scoring functions that are not tailored to this target class. Due to the open-source and modular architecture of ParaDockS further scoring funtions can be easily implemented and immediately used for docking and virtual screening. The whole ParaDockS suite is distributed under the terms of the GNU General Public License Version 2 and is free to download, to reuse, to modify and to redistribute any file of the source code.

Improving the Accuracy of Knowledge-Based Scoring Functions for Protein-Ligand Interactions by Accounting for Sparse Data in the Training Set

In the derivation of any knowledge-based scoring function, one must decide how to manage sparse features in the training data. Here we present STScore, a distance-dependent set of atomic pair potentials that uses a novel approach to minimize the sparse data problem. The overall approach is to represent the actual potential of mean force, which is unknown, as a random variable whose probability density is determined by the evidence in the training set. This provides a natural way of representing the uncertainty in the potential of mean force (PMF), for each bin and atom pair type. STScore is an average of the PMF and an alternative force-field-based potential, with the weights chosen to minimize the error in the sum. This weighting scheme implies that STScore will give more weight to the force-field potential whenever training data is scarce. We show that STScore effectively combines the two alternatives, exceeding the performance of either potential alone, and leads to improved binding mode and binding affinity predictions.

Prediction of the Monomer-Monomer Interface Region of the ß2Ar Homodimer Receptor via Docking Experiments

Increasing body of evidence indicate that G protein-coupled receptors (GPCRs) exist as dimers and oligomers in the cell membrane. The goal of this study is to estimate the homodimeric form of beta-2 adrenergic (s2AR) receptor which is a member of the GPCR family, via multiple docking experiments.The transmembrane domain VI (TMVI) of the receptor is suggested to be a significant part of the interface both experimentally and theoretically [1,2]. The amino acid sequence motif, 75LIXXGVXXG83VXXT, which is proposed to be essential for dimerization in glycophorin A is similarly observed on TMVI of s2AR as in LKTLG276IMMG280TFTL284. A peptide is derived from TMVI consisting of residues from 276 to 296, GIIMGTFTLCWLPFFIVNIVH and blindly docked to one conformer of the monomeric structure using a rigid body approach, via AutoDock v4.0 software tool [3]. Bound conformations were then reevaluated with a knowledge-based scoring function called DSXonline v0.88 [4].Docking results shows that the peptide has the highest binding affinity for TMVI which supports the possible role of TMVI at the interface of a dimeric structure. Also, a residue interface propensity data derived from a set of 32 nonhomologous homodimers [5] is able to identify TMVI as the domain which contains the highest number of residues with the highest interface propensity values.