MedusaScore: an accurate force field-based scoring function for virtual drug screening.

We report herein our efforts in the development of three empirical scoring functions with application in protein-ligand docking. A first scoring function was developed from 209 crystal structures of protein-ligand complexes and a second one from 946 cross-docked complexes. Tuning of the coefficients for the different terms making up these functions was performed by an iterative approach to optimize the correlations between observed activities and calculated scores. A third scoring function was developed from libraries of known actives and decoys docked to six different protein conformational ensembles. In the latter case, the tuning of the coefficients was performed so as to optimize the area under the curve of a receiver operating characteristic (ROC) for the discrimination of actives and inactives. The newly developed scoring functions were next assessed on independent sets of protein-ligand complexes for their ability to predict binding affinities and to discriminate actives from inactives. In the first validation the first function, which was trained on active compounds only, performed as well as other commonly used ones. On a high-throughput virtual screening validation on five protein conformational ensembles, the third scoring function that included data from inactive compounds performed significantly better. This validation showed that the inclusion of data from inactive compounds is critical for performance in virtual high-throughput screening applications.

Computational modeling of protein-ligand interactions plays a crucial role in drug discovery by enabling the prediction of binding affinities, identification of potential drug candidates, and optimization of lead compounds. This approach leverages molecular docking, molecular dynamics simulations, and other computational techniques to understand how small molecules interact with target proteins. In this article, we review the key computational methods used to model protein-ligand interactions, including their applications in virtual screening, drug optimization, and mechanistic studies. We also discuss the challenges and future directions in computational drug discovery, emphasizing the role of artificial intelligence and machine learning in enhancing predictive models.

Identification of HIV-1 reverse transcriptase dual inhibitors by a combined shape-, 2D-fingerprint- and pharmacophore-based virtual screening approach

The holistic integration of virtual screening in drug discovery

Virtual screening can be considered as anin silicoanalog to high throughput screening. Its goal is the drastic reduction of compound libraries to a manageable size for synthesis and biological testing. Virtual screening applications are not standardized and packaged in single software products. They rather consist of cascades of diverse filter programs that address different characteristics of putative lead chemical matter. Fast counting schemes for profiling physicochemical properties and machine learning approaches address druglike properties of small molecules related to pharmacology and toxicology. Focused libraries for target classes can be designed, for example, for combinatorial synthesis. Pharmacophores derived from known ligands of a given target can be used in virtual pharmacophore screening. Finally, structure‐based virtual screening employs docking and scoring approaches to find the binding mode of putative protein‐ligand complexes and estimate their binding affinities.

The use of virtual screening and differential scanning fluorimetry for the rapid identification of fragments active against MEK1

It has been reported recently that consensus scoring, which combines multiple scoring functions in binding affinity estimation, leads to higher hit-rates in virtual library screening studies. This method seems quite independent to the target receptor, the docking program, or even the scoring functions under investigation. Here we present an idealized computer experiment to explore how consensus scoring works. A hypothetical set of 5000 compounds is used to represent a chemical library under screening. The binding affinities of all its member compounds are assigned by mimicking a real situation. Based on the assumption that the error of a scoring function is a random number in a normal distribution, the predicted binding affinities were generated by adding such a random number to the "observed" binding affinities. The relationship between the hit-rates and the number of scoring functions employed in scoring was then investigated. The performance of several typical ranking strategies for a consensus scoring procedure was also explored. Our results demonstrate that consensus scoring outperforms any single scoring for a simple statistical reason: the mean value of repeated samplings tends to be closer to the true value. Our results also suggest that a moderate number of scoring functions, three or four, are sufficient for the purpose of consensus scoring. As for the ranking strategy, both the rank-by-number and the rank-by-rank strategy work more effectively than the rank-by-vote strategy.

Powered by dramatic advances in computer hardware, the advent of ultralarge make-on-demand virtual libraries, and a shift in small-molecule discovery toward more challenging targets with limited known actives, there has been a growing interest in the development of performant virtual screening methods that can reliably deliver novel hits. We report on a new method called Glide WS, that builds on our earlier efforts (WScore) to introduce an explicit representation of water structure and dynamics to an empirical scoring function suitable for high-throughput docking. This scoring function has been carefully tuned using absolute binding free energy perturbation calculations (ABFEP). Compared with Glide SP, Glide WS offers significant gains in the two primary tasks for molecular docking in drug discovery, pose prediction and virtual screening enrichment. For docking accuracy, Glide WS achieves a self-docking accuracy of 92% on a diverse set of 1477 protein ligand complexes as compared to 85% for Glide SP, using a criterion of 2.5 Å. We also demonstrate significantly improved virtual screening enrichment using a diverse data set covering of 38 targets together with three different computationally generated libraries of decoys, combined with standard known ChEMBL actives. We focus on ligands ranked in the top few percent of the database (the subset that is relevant to practical virtual screening efforts) and demonstrate that, along with improved enrichment of ChEMBL actives, Glide WS achieves a remarkable reduction in the number of poorly scoring decoys (as calibrated by ABFEP calculations), across a high percentage of targets, as compared to Glide SP. These results suggest that considerably higher hit rates will be observed, as compared to conventional rigid receptor docking, in practical virtual screening applications.

A high throughput framework for virtual screening (VS) applications has been designed which enables researchers in in-silico drug discovery to deploy large scale time-consuming virtual screening experiments on High Performance/Grid computing environments. Issues related to enabling virtual screening on grids have been addressed extensively in the past; in this paper we describe a new framework, which supports the rapid deployment of virtual screening services based on UNICORE-6 and Meta Scheduling Service (MSS). We demonstrate, how the deployment of complex virtual screening workflows is facilitated by this framework and we provide some first results of the virtual laboratory testbed established in the course of the EU project PHOSPHORUS. The flexibility of the new VS framework allows easy deployment and integration of new resources and applications, making the Grid more accessible to the users.

An accurate energy scoring function is crucial for protein structure prediction. Given the increasing number of experimentally determined structures, knowledge-based approaches have been widely used to develop scoring functions for protein structure prediction in the past three decades. However, current scoring functions often only consider nonbonded interactions and neglect bonded potentials like covalent bonds and angles for the sake of speed and simplicity. Although such scoring functions may be successful on fully relaxed conformations, they would have difficulties in ranking those decoys with distorted bonds or angles, especially when being used for conformational sampling in structure prediction. Therefore, such a scoring function may perform well on one or several decoy sets, but it often has limited accuracy on large diverse sets. Addressing the limitation, we have developed a composite knowledge-based scoring function, named as ITCPS, by integrating bonded and nonbonded potentials as well as orientation-dependent and hydrophobic interactions. Our scoring function ITCPS was extensively evaluated on 18 decoy sets of 927 proteins including three sets of 3DRobot, AMBER benchmarking set, HR, CASP5-8, CASP9-13, eight sets of Decoy 'R' Us, MOULDER, ROSETTA, and I-TASSER set and compared with 51 other scoring functions. It was shown that overall ITCPS performed the best among the 52 scoring functions and achieved a good performance on all the test sets. Of 927 proteins, ITCPS recognized the native structures for 842 proteins, giving a success rate of 90.8% and an average Z-score of 3.36. Moreover, ITCPS also exhibited a strong ability to distinguish the best near-native structure among decoys and achieved a significantly better performance than other tested scoring functions. The present model is expected to be beneficial for the development of scoring functions for other interactions.

Machine-learning (ML)-based scoring functions (MLSFs) have gradually emerged as a promising alternative for protein-ligand binding affinity prediction and structure-based virtual screening. However, clouds of doubts have still been raised against the benefits of this novel type of scoring functions (SFs). In this study, to benchmark the performance of target-specific MLSFs on a relatively unbiased dataset, the MLSFs trained from three representative protein-ligand interaction representations were assessed on the LIT-PCBA dataset, and the classical Glide SP SF and three types of ligand-based quantitative structure-activity relationship (QSAR) models were also utilized for comparison. Two major aspects in virtual screening campaigns, including prediction accuracy and hit novelty, were systematically explored. The calculation results illustrate that the tested target-specific MLSFs yielded generally superior performance over the classical Glide SP SF, but they could hardly outperform the 2D fingerprint-based QSAR models. Although substantial improvements could be achieved by integrating multiple types of protein-ligand interaction features, the MLSFs were still not sufficient to exceed MACCS-based QSAR models. In terms of the correlations between the hit ranks or the structures of the top-ranked hits, the MLSFs developed by different featurization strategies would have the ability to identify quite different hits. Nevertheless, it seems that target-specific MLSFs do not have the intrinsic attributes of a traditional SF and may not be a substitute for classical SFs. In contrast, MLSFs can be regarded as a new derivative of ligand-based QSAR models. It is expected that our study may provide valuable guidance for the assessment and further development of target-specific MLSFs.

Structure-based virtual screening techniques require reliable scoring functions to discriminate potential substrates effectively. In this study we compared the performance of GOLD, PMF, DOCK and FlexX scoring functions in FlexX flexible docking to cytochrome P450cam binding site. Crystal structures of protein-substrate complexes were most effectively reproduced by the FlexX/PMF method. On the other hand, the FlexX/GOLD approach provided the best correlation between experimental binding constants and predicted scores. Binding modes selected by the FlexX/PMF approach were rescored by GOLD to obtain a reliable measure of binding energetics. The effectiveness of the FlexX/PMF/GOLD method was demonstrated by the correct classification of 32 out of the 33 experimentally studied compounds and also in a virtual HTS test on a library of 10,000 compounds. Although almost all the available functions were developed to be general, our study on cytochrome P450cam substrates suggests that careful selection or even tailoring the scoring function might increase the prediction power of virtual screens significantly. The FlexX/PMF/GOLD methodology was tested on cytochrome P450 3A4 substrates and inhibitors. This preliminary study revealed that the combined function was able to recognise 334 out of the 345 compounds bound to 3A4.

Virtual Screening (VS) based on molecular docking is an efficient method used for retrieving novel hit compounds in drug discovery. However, the accuracy of the current docking scoring function (SF) is usually insufficient. In this study, in order to improve the screening power of SF, a novel approach named EAT-Score was proposed by directly utilizing the energy auxiliary terms (EAT) provided by molecular docking scoring through eXtreme Gradient Boosting (XGBoost). Here, EAT specifically refers to the output of the Molecular Operating Environment (MOE) scoring, including the energy scores of five different classical SFs and the Protein-Ligand Interaction Fingerprint (PLIF) terms. The performance of EAT-Score to discriminate actives from decoys was strictly validated on the DUD-E diverse subset by using different performance metrics. The results showed that EAT-Score performed much better than classical SFs in VS, with its AUC values exhibiting an improvement of around 0.3. Meanwhile, EAT-Score could achieve comparable even better prediction performance compared with other state-of-the-art VS methods, such as some machine learning (ML)-based SFs and classical SFs implemented in docking programs, in terms of AUC, LogAUC, or BEDROC. Furthermore, the EAT-Score model can capture important binding pattern information from protein-ligand complexes by Shapley additive explanations (SHAP) analysis, which may be very helpful in interpreting the ligand binding mechanism for a certain target and thereby guiding drug design.

In recent years, machine learning has been proposed as a promising strategy to build accurate scoring functions for computational docking finalized to numerically empowered drug discovery. However, the latest studies have suggested that over-optimistic results had been reported due to the correlations present in the experimental databases used for training and testing. Here, we investigate the performance of an artificial neural network in binding affinity predictions, comparing results obtained using both experimental protein-ligand structures as well as larger sets of computer-generated structures created using commercial software. Interestingly, similar performances are obtained on both databases. We find a noticeable performance suppression when moving from random horizontal tests to vertical tests performed on target proteins not included in the training data. The possibility to train the network on relatively easily created computer-generated databases leads us to explore per-target scoring functions, trained and tested ad-hoc on complexes including only one target protein. Encouraging results are obtained, depending on the type of protein being addressed.

Modulation of autotaxin (ATX), the lysophospholipase D enzyme that produces lysophosphatidic acid, with small-molecule inhibitors is a promising strategy for blocking the ATX-lysophosphatidic acid signaling axis. Although discovery campaigns have been successful in identifying ATX inhibitors, many of the reported inhibitors target the catalytic cleft of ATX. A recent study provided evidence for an additional inhibitory surface in the hydrophobic binding pocket of ATX, confirming prior studies that relied on enzyme kinetics and differential inhibition of substrates varying in size. Multiple hits from previous high-throughput screening for ATX inhibitors were obtained with aromatic sulfonamide derivatives interacting with the hydrophobic pocket. Here, we describe the development of a ligand-based strategy and its application in virtual screening, which yielded novel high-potency inhibitors that target the hydrophobic pocket of ATX. Characterization of the structure-activity relationship of these new inhibitors forms the foundation of a new pharmacophore model of the hydrophobic pocket of ATX.