Effects of inductive bias on computational evaluations of ligand-based modeling and on drug discovery

DrugGym: A testbed for the economics of autonomous drug discovery

Drug discovery is stochastic. The effectiveness of candidate compounds in satisfying design objectives is unknown ahead of time, and the tools used for prioritization-predictive models and assays-are inaccurate and noisy. In a typical discovery campaign, thousands of compounds may be synthesized and tested before design objectives are achieved, with many others ideated but deprioritized. These challenges are well-documented, but assessing potential remedies has been difficult. We introduce DrugGym, a framework for modeling the stochastic process of drug discovery. Emulating biochemical assays with realistic surrogate models, we simulate the progression from weak hits to sub-micromolar leads with viable ADME. We use this testbed to examine how different ideation, scoring, and decision-making strategies impact statistical measures of utility, such as the probability of program success within predefined budgets and the expected costs to achieve target candidate profile (TCP) goals. We also assess the influence of affinity model inaccuracy, chemical creativity, batch size, and multi-step reasoning. Our findings suggest that reducing affinity model inaccuracy from 2 to 0.5 pIC50 units improves budget-constrained success rates tenfold. DrugGym represents a realistic testbed for machine learning methods applied to the hit-to-lead phase. Source code is available at www.drug-gym.org.

Advances in machine intelligence-driven virtual screening approaches for big-data

Virtual screening (VS) is an integral and ever-evolving domain of drug discovery framework. The VS is traditionally classified into ligand-based (LB) and structure-based (SB) approaches. Machine intelligence or artificial intelligence has wide applications in the drug discovery domain to reduce time and resource consumption. In combination with machine intelligence algorithms, VS has emerged into revolutionarily progressive technology that learns within robust decision orders for data curation and hit molecule screening from large VS libraries in minutes or hours. The exponential growth of chemical and biological data has evolved as "big-data" in the public domain demands modern and advanced machine intelligence-driven VS approaches to screen hit molecules from ultra-large VS libraries. VS has evolved from an individual approach (LB and SB) to integrated LB and SB techniques to explore various ligand and target protein aspects for the enhanced rate of appropriate hit molecule prediction. Current trends demand advanced and intelligent solutions to handle enormous data in drug discovery domain for screening and optimizing hits or lead with fewer or no false positive hits. Following the big-data drift and tremendous growth in computational architecture, we presented this review. Here, the article categorized and emphasized individual VS techniques, detailed literature presented for machine learning implementation, modern machine intelligence approaches, and limitations and deliberated the future prospects.

Deep reinforcement learning enables better bias control in benchmark for virtual screening

Virtual screening (VS) has been incorporated into the paradigm of modern drug discovery. This field is now undergoing a new wave of revolution driven by artificial intelligence and more specifically, machine learning (ML). In terms of those out-of-the-box datasets for model training or benchmarking, their data volume and applicability domain are limited. They are suffering from the biases constantly reported in the ML application. To address these issues, we present a novel benchmark named MUBDsyn. The utilization of synthetic decoys (i.e., presumed inactives) is the main feature of MUBDsyn, where deep reinforcement learning was leveraged for bias control during decoy generation. Then, we carried out extensive validations on this new benchmark. First, we confirmed that MUBDsyn was superior to the classical benchmarks in control of domain bias, artificial enrichment bias and analogue bias. Moreover, we found that the assessment of ML models based on MUBDsyn was less biased as revealed by the analysis of asymmetric validation embedding bias. In addition, MUBDsyn showed better setting of benchmarking challenge for deep learning models compared with NRLiSt-BDB. Overall, we have proven that MUBDsyn is the close-to-ideal benchmark for VS. The computational tool is publicly available for the easy extension of MUBDsyn.

Combatting over-specialization bias in growing chemical databases

BACKGROUND

Predicting in advance the behavior of new chemical compounds can support the design process of new products by directing the research toward the most promising candidates and ruling out others. Such predictive models can be data-driven using Machine Learning or based on researchers' experience and depend on the collection of past results. In either case: models (or researchers) can only make reliable assumptions about compounds that are similar to what they have seen before. Therefore, consequent usage of these predictive models shapes the dataset and causes a continuous specialization shrinking the applicability domain of all trained models on this dataset in the future, and increasingly harming model-based exploration of the space.

PROPOSED SOLUTION

In this paper, we propose CANCELS (CounterActiNg Compound spEciaLization biaS), a technique that helps to break the dataset specialization spiral. Aiming for a smooth distribution of the compounds in the dataset, we identify areas in the space that fall short and suggest additional experiments that help bridge the gap. Thereby, we generally improve the dataset quality in an entirely unsupervised manner and create awareness of potential flaws in the data. CANCELS does not aim to cover the entire compound space and hence retains a desirable degree of specialization to a specified research domain.

RESULTS

An extensive set of experiments on the use-case of biodegradation pathway prediction not only reveals that the bias spiral can indeed be observed but also that CANCELS produces meaningful results. Additionally, we demonstrate that mitigating the observed bias is crucial as it cannot only intervene with the continuous specialization process, but also significantly improves a predictor's performance while reducing the number of required experiments. Overall, we believe that CANCELS can support researchers in their experimentation process to not only better understand their data and potential flaws, but also to grow the dataset in a sustainable way. All code is available under github.com/KatDost/Cancels .

Computer Science Technology in Natural Products Research: A Review of Its Applications and Implications

Computational approaches to drug development are rapidly growing in popularity and have been used to produce significant results. Recent developments in information science have expanded databases and chemical informatics knowledge relating to natural products. Natural products have long been well-studied, and a large number of unique structures and remarkable active substances have been reported. Analyzing accumulated natural product knowledge using emerging computational science techniques is expected to yield more new discoveries. In this article, we discuss the current state of natural product research using machine learning. The basic concepts and frameworks of machine learning are summarized. Natural product research that utilizes machine learning is described in terms of the exploration of active compounds, automatic compound design, and application to spectral data. In addition, efforts to develop drugs for intractable diseases will be addressed. Lastly, we discuss key considerations for applying machine learning in this field. This paper aims to promote progress in natural product research by presenting the current state of computational science and chemoinformatics approaches in terms of its applications, strengths, limitations, and implications for the field.

MILCDock: Machine Learning Enhanced Consensus Docking for Virtual Screening in Drug Discovery

Molecular docking tools are regularly used to computationally identify new molecules in virtual screening for drug discovery. However, docking tools suffer from inaccurate scoring functions with widely varying performance on different proteins. To enable more accurate ranking of active over inactive ligands in virtual screening, we created a machine learning consensus docking tool, MILCDock, that uses predictions from five traditional molecular docking tools to predict the probability a ligand binds to a protein. MILCDock was trained and tested on data from both the DUD-E and LIT-PCBA docking datasets and shows improved performance over traditional molecular docking tools and other consensus docking methods on the DUD-E dataset. LIT-PCBA targets proved to be difficult for all methods tested. We also find that DUD-E data, although biased, can be effective in training machine learning tools if care is taken to avoid DUD-E's biases during training.

Complex machine learning model needs complex testing: Examining predictability of molecular binding affinity by a graph neural network

Drug discovery pipelines typically involve high-throughput screening of large amounts of compounds in a search of potential drugs candidates. As a chemical space of small organic molecules is huge, a "navigation" over it urges for fast and lightweight computational methods, thus promoting machine-learning approaches for processing huge pools of candidates. In this contribution, we present a graph-based deep neural network for prediction of protein-drug binding affinity and assess its predictive power under thorough testing conditions. Within the suggested approach, both protein and drug molecules are represented as graphs and passed to separate graph sub-networks, then concatenated and regressed towards a binding affinity. The neural network is trained on two binding affinity datasets-PDBbind and data imported from RCSB Protein Data Bank. In order to explore the generalization capabilities of the model we go beyond traditional random or leave-cluster-out techniques and demonstrate the need for more elaborate model performance assessment - six different strategies for test/train data partitioning (random, time- and property-arranged, protein- and ligand-clustered) with a k-fold cross-validation are engaged. Finally, we discuss the model performance in terms of a set of metrics for different split strategies and fold arrangement. Our code is available at https://github.com/SoftServeInc/affinity-by-GNN.

Synergy and Complementarity between Focused Machine Learning and Physics-Based Simulation in Affinity Prediction

We present results on the extent to which physics-based simulation (exemplified by FEP⁺) and focused machine learning (exemplified by QuanSA) are complementary for ligand affinity prediction. For both methods, predictions of activity for LFA-1 inhibitors from a medicinal chemistry lead optimization project were accurate within the applicable domain of each approach. A hybrid model that combined predictions by both approaches by simple averaging performed better than either method, with respect to both ranking and absolute pK_i values. Two publicly available FEP⁺ benchmarks, covering 16 diverse biological targets, were used to test the generality of the synergy. By identifying training data specifically focused on relevant ligands, accurate QuanSA models were derived using ligand activity data known at the time of the original series publications. Results across the 16 benchmark targets demonstrated significant improvements both for ranking and for absolute pK_i values using hybrid predictions that combined the FEP⁺ and QuanSA predicted affinity values. The results argue for a combined approach for affinity prediction that makes use of physics-driven methods as well as those driven by machine learning, each applied carefully on appropriate compounds, with hybrid prediction strategies being employed where possible.

Comparison of structure- and ligand-based scoring functions for deep generative models: a GPCR case study

Deep generative models have shown the ability to devise both valid and novel chemistry, which could significantly accelerate the identification of bioactive compounds. Many current models, however, use molecular descriptors or ligand-based predictive methods to guide molecule generation towards a desirable property space. This restricts their application to relatively data-rich targets, neglecting those where little data is available to sufficiently train a predictor. Moreover, ligand-based approaches often bias molecule generation towards previously established chemical space, thereby limiting their ability to identify truly novel chemotypes. In this work, we assess the ability of using molecular docking via Glide-a structure-based approach-as a scoring function to guide the deep generative model REINVENT and compare model performance and behaviour to a ligand-based scoring function. Additionally, we modify the previously published MOSES benchmarking dataset to remove any induced bias towards non-protonatable groups. We also propose a new metric to measure dataset diversity, which is less confounded by the distribution of heavy atom count than the commonly used internal diversity metric. With respect to the main findings, we found that when optimizing the docking score against DRD2, the model improves predicted ligand affinity beyond that of known DRD2 active molecules. In addition, generated molecules occupy complementary chemical and physicochemical space compared to the ligand-based approach, and novel physicochemical space compared to known DRD2 active molecules. Furthermore, the structure-based approach learns to generate molecules that satisfy crucial residue interactions, which is information only available when taking protein structure into account. Overall, this work demonstrates the advantage of using molecular docking to guide de novo molecule generation over ligand-based predictors with respect to predicted affinity, novelty, and the ability to identify key interactions between ligand and protein target. Practically, this approach has applications in early hit generation campaigns to enrich a virtual library towards a particular target, and also in novelty-focused projects, where de novo molecule generation either has no prior ligand knowledge available or should not be biased by it.

Improved 3D-QSAR Prediction by Multiple Conformational Alignments and Molecular Docking Studies to Design and Discover HIV-I Protease Inhibitors

BACKGROUND

Inhibition of HIV-I protease enzyme is a strategic step for providing better treatment in retrovirus infections, which avoids resistance and possesses less toxicity.

OBJECTIVES

In the course of our research to discover new and potent protease inhibitors, 3D-QSAR (CoMFA and CoMSIA) models were generated using 3 different alignment techniques, including multifit alignment, docking based and Distill based alignment for 63 compounds. Novel molecules were designed from the output of this study.

METHODS

A total of 3 alignment methods were used to generate CoMFA and CoMSIA models. A Distill based alignment method was considered a better method according to different validation parameters. A 3D-QSAR model was generated and contour maps were discussed. The biological activity of designed molecules was predicted using the generated QSAR model to validate QSAR. The newly designed molecules were docked to predict binding affinity.

RESULTS

In CoMFA, leave one out cross-validated coefficient (q²), conventional coefficient (r²) and predicted correlation coefficient (r²_Predicted) values were found to be 0.721, 0.991 and 0.780, respectively. The best obtained CoMSIA model also showed significant cross-validated coefficient (q²), conventional coefficient (r²) and predicted correlation coefficient (r²_Predicted) values of 0.714, 0.987 and 0.721, respectively. Steric and electrostatic contour maps generated from CoMFA and hydrophobic and hydrogen bond donor and hydrogen bond acceptor contour maps from CoMSIA models were used to design new and bioactive protease inhibitors by incorporating bioisosterism and knowledge-based structure-activity relationship.

CONCLUSION

The results from both these approaches, ligand-based drug design and structure-based drug design, are adequate and promising to discover protease inhibitors.

Structure- and Ligand-Based Virtual Screening on DUD-E+: Performance Dependence on Approximations to the Binding Pocket

Using the DUD-E⁺ benchmark, we explore the impact of using a single protein pocket or ligand for virtual screening compared with using ensembles of alternative pockets, ligands, and sets thereof. For both structure-based and ligand-based approaches, the precise characterization of the binding site in question had a significant impact on screening performance. Using the single original DUD-E protein, Surflex-Dock yielded mean ROC area of 0.81 ± 0.11. Using the cognate ligand instead, with the eSim method for screening, yielded 0.77 ± 0.14. Moving to ensembles of five protein pocket variants increased docking performance to 0.84 ± 0.09. Results for the analogous ligand-based approach (using the five crystallographically aligned cognate ligands) was 0.83 ± 0.11. Using the same ligands, but making use of an automatically generated mutual alignment, yielded mean AUC nearly as good as from single-structure docking: 0.80 ± 0.12. Detailed results and statistical analyses show that structure- and ligand-based methods are complementary and can be fruitfully combined to enhance screening efficiency. A hybrid approach combining ensemble docking with eSim-based screening produced the best and most consistent performance (mean ROC area of 0.89 ± 0.08 and 1% early enrichment of 46-fold). Based on results from both the docking and ligand-similarity approaches, it is clearly unwise to make use of a single arbitrarily chosen protein structure for docking or single ligand query for similarity-based screening.

The Effect of Debiasing Protein-Ligand Binding Data on Generalization

The structured nature of chemical data means machine-learning models trained to predict protein-ligand binding risk overfitting the data, impairing their ability to generalize and make accurate predictions for novel candidate ligands. Data debiasing algorithms, which systematically partition the data to reduce bias and provide a more accurate metric of model performance, have the potential to address this issue. When models are trained using debiased data splits, the reward for simply memorizing the training data is reduced, suggesting that the ability of the model to make accurate predictions for novel candidate ligands will improve. To test this hypothesis, we use distance-based data splits to measure how well a model can generalize. We first confirm that models perform better for randomly split held-out sets than for distant held-out sets. We then debias the data and find, surprisingly, that debiasing typically reduces the ability of models to make accurate predictions for distant held-out test sets and that model performance measured after debiasing is not representative of the ability of a model to generalize. These results suggest that debiasing reduces the information available to a model, impairing its ability to generalize.

Current methods and challenges for deep learning in drug discovery

Driven by rapid advances in computer hardware and publicly available datasets over the past decade, deep learning has achieved tremendous success in the transformation of many computational disciplines. These novel technologies have had considerable impact on computer-aided drug design as well, throughout all stages of the development pipeline. A flexible toolbox of neural architectures has been developed that are well-suited to represent the sequential, topological, or geometrical concepts of chemistry and biology; and that are able to either discriminate existing molecules or to generate new ones from scratch. For some biochemical prediction tasks, the state of the art has been advanced; however, for complex and practically relevant projects, the outcomes are less clear-cut. Current deep learning methods rely on massive amounts of labeled examples, but drug discovery data is comparatively limited in quantity and quality. These problems need to be resolved and existing sources used more effectively to demonstrate that deep learning can revolutionize the field in general.

Development of New Methods Needs Proper Evaluation-Benchmarking Sets for Machine Learning Experiments for Class A GPCRs

New computational approaches for virtual screening applications are constantly being developed. However, before a particular tool is used to search for new active compounds, its effectiveness in the type of task must be examined. In this study, we conducted a detailed analysis of various aspects of preparation of respective data sets for such an evaluation. We propose a protocol for fetching data from the ChEMBL database, examine various compound representations in terms of the possible bias resulting from the way they are generated, and define a new metric for comparing the structural similarity of compounds, which is in line with chemical intuition. The newly developed method is also used for the evaluation of various approaches for division of the data set into training and test set parts, which are also examined in detail in terms of being the source of possible results bias. Finally, machine learning methods are applied in cross-validation studies of data sets constructed within the paper, constituting benchmarks for the assessment of computational methods developed for virtual screening tasks. Additionally, analogous data sets for class A G protein-coupled receptors (100 targets with the highest number of records) were prepared. They are available at http://gmum.net/benchmarks/ , together with script enabling reproduction of all results available at https://github.com/lesniak43/ananas .

Electrostatic-field and surface-shape similarity for virtual screening and pose prediction

We introduce a new method for rapid computation of 3D molecular similarity that combines electrostatic field comparison with comparison of molecular surface-shape and directional hydrogen-bonding preferences (called "eSim"). Rather than employing heuristic "colors" or user-defined molecular feature types to represent conformation-dependent molecular electrostatics, eSim calculates the similarity of the electrostatic fields of two molecules (in addition to shape and hydrogen-bonding). We present detailed virtual screening performance data on the standard 102 target DUD-E set. In its moderately fast screening mode, eSim running on a single computing core is capable of processing over 60 molecules per second. In this mode, eSim performed significantly better than all alternate methods for which full DUD-E data were available (mean ROC area of 0.74, p [Formula: see text], by paired t-test, compared with the best performing alternate method). In addition, for 92 targets of the DUD-E set where multiple ligand-bound crystal structures were available, screening performance was assessed using alternate ligands or sets thereof (in their bound poses) as similarity targets. Using the joint alignment of five ligands for each protein target, mean ROC area exceeded 0.82 for the 92 targets. Design-focused application of ligand similarity methods depends on accurate predictions of geometric molecular relationships. We comprehensively assessed pose prediction accuracy by curating nearly 400,000 bound ligand pose pairs across the DUD-E targets. Overall, beginning from agnostic initial poses, we observed an 80% success rate for RMSD [Formula: see text] Å  among the top 20 predicted eSim poses. These examples were split roughly 50/50 into cases with high direct atomic overlap (where a shared scaffold exists between a pair) and low direct atomic overlap (where where a ligand pair has dissimilar scaffolds but largely occupies the same space). Within the high direct atomic overlap subset, the pose prediction success rate was 93%. For the more challenging subset (where dissimilar scaffolds are to be aligned), the success rate was 70%. The eSim approach enables both large-scale screening and rational design of ligands and is rooted in physically meaningful, non-heuristic, molecular comparisons.

Anthropogenic biases in chemical reaction data hinder exploratory inorganic synthesis

Most chemical experiments are planned by human scientists and therefore are subject to a variety of human cognitive biases¹, heuristics² and social influences³. These anthropogenic chemical reaction data are widely used to train machine-learning models⁴ that are used to predict organic⁵ and inorganic^6,7 syntheses. However, it is known that societal biases are encoded in datasets and are perpetuated in machine-learning models⁸. Here we identify as-yet-unacknowledged anthropogenic biases in both the reagent choices and reaction conditions of chemical reaction datasets using a combination of data mining and experiments. We find that the amine choices in the reported crystal structures of hydrothermal synthesis of amine-templated metal oxides⁹ follow a power-law distribution in which 17% of amine reactants occur in 79% of reported compounds, consistent with distributions in social influence models^10-12. An analysis of unpublished historical laboratory notebook records shows similarly biased distributions of reaction condition choices. By performing 548 randomly generated experiments, we demonstrate that the popularity of reactants or the choices of reaction conditions are uncorrelated to the success of the reaction. We show that randomly generated experiments better illustrate the range of parameter choices that are compatible with crystal formation. Machine-learning models that we train on a smaller randomized reaction dataset outperform models trained on larger human-selected reaction datasets, demonstrating the importance of identifying and addressing anthropogenic biases in scientific data.

In Need of Bias Control: Evaluating Chemical Data for Machine Learning in Structure-Based Virtual Screening

Reports of successful applications of machine learning (ML) methods in structure-based virtual screening (SBVS) are increasing. ML methods such as convolutional neural networks show promising results and often outperform traditional methods such as empirical scoring functions in retrospective validation. However, trained ML models are often treated as black boxes and are not straightforwardly interpretable. In most cases, it is unknown which features in the data are decisive and whether a model's predictions are right for the right reason. Hence, we re-evaluated three widely used benchmark data sets in the context of ML methods and came to the conclusion that not every benchmark data set is suitable. Moreover, we demonstrate on two examples from current literature that bias is learned implicitly and unnoticed from standard benchmarks. On the basis of these results, we conclude that there is a need for eligible validation experiments and benchmark data sets suited to ML for more bias-controlled validation in ML-based SBVS. Therefore, we provide guidelines for setting up validation experiments and give a perspective on how new data sets could be generated.

Quantitative surface field analysis: learning causal models to predict ligand binding affinity and pose

We introduce the QuanSA method for inducing physically meaningful field-based models of ligand binding pockets based on structure-activity data alone. The method is closely related to the QMOD approach, substituting a learned scoring field for a pocket constructed of molecular fragments. The problem of mutual ligand alignment is addressed in a general way, and optimal model parameters and ligand poses are identified through multiple-instance machine learning. We provide algorithmic details along with performance results on sixteen structure-activity data sets covering many pharmaceutically relevant targets. In particular, we show how models initially induced from small data sets can extrapolatively identify potent new ligands with novel underlying scaffolds with very high specificity. Further, we show that combining predictions from QuanSA models with those from physics-based simulation approaches is synergistic. QuanSA predictions yield binding affinities, explicit estimates of ligand strain, associated ligand pose families, and estimates of structural novelty and confidence. The method is applicable for fine-grained lead optimization as well as potent new lead identification.

Most Ligand-Based Classification Benchmarks Reward Memorization Rather than Generalization

Undetected overfitting can occur when there are significant redundancies between training and validation data. We describe AVE, a new measure of training-validation redundancy for ligand-based classification problems, that accounts for the similarity among inactive molecules as well as active ones. We investigated seven widely used benchmarks for virtual screening and classification, and we show that the amount of AVE bias strongly correlates with the performance of ligand-based predictive methods irrespective of the predicted property, chemical fingerprint, similarity measure, or previously applied unbiasing techniques. Therefore, it may be the case that the previously reported performance of most ligand-based methods can be explained by overfitting to benchmarks rather than good prospective accuracy.

Maximal Unbiased Benchmarking Data Sets for Human Chemokine Receptors and Comparative Analysis

Chemokine receptors (CRs) have long been druggable targets for the treatment of inflammatory diseases and HIV-1 infection. As a powerful technique, virtual screening (VS) has been widely applied to identifying small molecule leads for modern drug targets including CRs. For rational selection of a wide variety of VS approaches, ligand enrichment assessment based on a benchmarking data set has become an indispensable practice. However, the lack of versatile benchmarking sets for the whole CRs family that are able to unbiasedly evaluate every single approach including both structure- and ligand-based VS somewhat hinders modern drug discovery efforts. To address this issue, we constructed Maximal Unbiased Benchmarking Data sets for human Chemokine Receptors (MUBD-hCRs) using our recently developed tools of MUBD-DecoyMaker. The MUBD-hCRs encompasses 13 subtypes out of 20 chemokine receptors, composed of 404 ligands and 15756 decoys so far and is readily expandable in the future. It had been thoroughly validated that MUBD-hCRs ligands are chemically diverse while its decoys are maximal unbiased in terms of "artificial enrichment", "analogue bias". In addition, we studied the performance of MUBD-hCRs, in particular CXCR4 and CCR5 data sets, in ligand enrichment assessments of both structure- and ligand-based VS approaches in comparison with other benchmarking data sets available in the public domain and demonstrated that MUBD-hCRs is very capable of designating the optimal VS approach. MUBD-hCRs is a unique and maximal unbiased benchmarking set that covers major CRs subtypes so far.

Statistical and machine learning approaches to predicting protein-ligand interactions

Data driven computational approaches to predicting protein-ligand binding are currently achieving unprecedented levels of accuracy on held-out test datasets. Up until now, however, this has not led to corresponding breakthroughs in our ability to design novel ligands for protein targets of interest. This review summarizes the current state of the art in this field, emphasizing the recent development of deep neural networks for predicting protein-ligand binding. We explain the major technical challenges that have caused difficulty with predicting novel ligands, including the problems of sampling noise and the challenge of using benchmark datasets that are sufficiently unbiased that they allow the model to extrapolate to new regimes.

Decoys Selection in Benchmarking Datasets: Overview and Perspectives

Virtual Screening (VS) is designed to prospectively help identifying potential hits, i.e., compounds capable of interacting with a given target and potentially modulate its activity, out of large compound collections. Among the variety of methodologies, it is crucial to select the protocol that is the most adapted to the query/target system under study and that yields the most reliable output. To this aim, the performance of VS methods is commonly evaluated and compared by computing their ability to retrieve active compounds in benchmarking datasets. The benchmarking datasets contain a subset of known active compounds together with a subset of decoys, i.e., assumed non-active molecules. The composition of both the active and the decoy compounds subsets is critical to limit the biases in the evaluation of the VS methods. In this review, we focus on the selection of decoy compounds that has considerably changed over the years, from randomly selected compounds to highly customized or experimentally validated negative compounds. We first outline the evolution of decoys selection in benchmarking databases as well as current benchmarking databases that tend to minimize the introduction of biases, and secondly, we propose recommendations for the selection and the design of benchmarking datasets.

A Systematic Review of Computational Drug Discovery, Development, and Repurposing for Ebola Virus Disease Treatment

Ebola virus disease (EVD) is a deadly global public health threat, with no currently approved treatments. Traditional drug discovery and development is too expensive and inefficient to react quickly to the threat. We review published research studies that utilize computational approaches to find or develop drugs that target the Ebola virus and synthesize its results. A variety of hypothesized and/or novel treatments are reported to have potential anti-Ebola activity. Approaches that utilize multi-targeting/polypharmacology have the most promise in treating EVD.

A Simple Representation of Three-Dimensional Molecular Structure

Statistical and machine learning approaches predict drug-to-target relationships from 2D small-molecule topology patterns. One might expect 3D information to improve these calculations. Here we apply the logic of the extended connectivity fingerprint (ECFP) to develop a rapid, alignment-invariant 3D representation of molecular conformers, the extended three-dimensional fingerprint (E3FP). By integrating E3FP with the similarity ensemble approach (SEA), we achieve higher precision-recall performance relative to SEA with ECFP on ChEMBL20 and equivalent receiver operating characteristic performance. We identify classes of molecules for which E3FP is a better predictor of similarity in bioactivity than is ECFP. Finally, we report novel drug-to-target binding predictions inaccessible by 2D fingerprints and confirm three of them experimentally with ligand efficiencies from 0.442-0.637 kcal/mol/heavy atom.

Validated UPLC-MS/MS method for the determination of tadalafil in human plasma and its application to a pharmacokinetic study

A simple, rapid, and reliable UPLC-MS/MS method was developed and validated for the determination of tadalafil in human plasma. The plasma samples were deproteinized with acetonitrile. Chromatographic separation was performed on a Shiseido C18 (100 × 2.1 mm, 2.7 µm) column with isocratic elution using 2.0 mM ammonium acetate and acetonitrile (55:45, v/v) with 0.1% formic acid at a flow rate of 0.7 mL/min. The total run time was 1 min per sample. The quantitative analysis was performed using multiple reaction monitoring at transition of m/z 390.4 → 268.3 for tadalafil and m/z 475.3 → 283.3 for sildenafil as an internal standard. The method was fully validated over a concentration range of 5–1,000 ng/mL with a lower quantification limit of 5 ng/mL. Intra- and inter-day precision (relative standard deviation, %RSD) were within 8.4% and accuracy (relative error, %RE) was lower than -3.2%. The developed and validated method was successfully applied to a pharmacokinetic study of tadalafil (20 mg) in Korean healthy male subjects (n = 12).

Computational analysis reveal inhibitory action of nimbin against dengue viral envelope protein

Dengue has emerged to be global health problem worldwide. Hence there is an immediate need to adopt new strategies in the development of effective anti-dengue drugs. Extracts from the leaves of Azadirachta indica has been traditionally used in folk medicine for viral infections. In the present study we report the anti-viral potency of nimbin, the active compound from the neem leaf extract against the envelope protein of dengue virus. Progression of viral entry into the host cell is facilitated by the envelope protein of dengue virus, suggesting; it as an effective anti-viral target. Nimbin is found to be effective against the envelope protein of all four types of dengue virus (dengue 1-4), which is evident from our in silico analysis. Our findings suggest the clinical importance of nimbin, which can serve as effective lead compound for further analysis.

Knowledge-guided docking: accurate prospective prediction of bound configurations of novel ligands using Surflex-Dock

Prediction of the bound configuration of small-molecule ligands that differ substantially from the cognate ligand of a protein co-crystal structure is much more challenging than re-docking the cognate ligand. Success rates for cross-docking in the range of 20-30 % are common. We present an approach that uses structural information known prior to a particular cutoff-date to make predictions on ligands whose bounds structures were determined later. The knowledge-guided docking protocol was tested on a set of ten protein targets using a total of 949 ligands. The benchmark data set, called PINC ("PINC Is Not Cognate"), is publicly available. Protein pocket similarity was used to choose representative structures for ensemble-docking. The docking protocol made use of known ligand poses prior to the cutoff-date, both to help guide the configurational search and to adjust the rank of predicted poses. Overall, the top-scoring pose family was correct over 60 % of the time, with the top-two pose families approaching a 75 % success rate. Correct poses among all those predicted were identified nearly 90 % of the time. The largest improvements came from the use of molecular similarity to improve ligand pose rankings and the strategy for identifying representative protein structures. With the exception of a single outlier target, the knowledge-guided docking protocol produced results matching the quality of cognate-ligand re-docking, but it did so on a very challenging temporally-segregated cross-docking benchmark.

Comparative modeling and benchmarking data sets for human histone deacetylases and sirtuin families

Histone deacetylases (HDACs) are an important class of drug targets for the treatment of cancers, neurodegenerative diseases, and other types of diseases. Virtual screening (VS) has become fairly effective approaches for drug discovery of novel and highly selective histone deacetylase inhibitors (HDACIs). To facilitate the process, we constructed maximal unbiased benchmarking data sets for HDACs (MUBD-HDACs) using our recently published methods that were originally developed for building unbiased benchmarking sets for ligand-based virtual screening (LBVS). The MUBD-HDACs cover all four classes including Class III (Sirtuins family) and 14 HDAC isoforms, composed of 631 inhibitors and 24609 unbiased decoys. Its ligand sets have been validated extensively as chemically diverse, while the decoy sets were shown to be property-matching with ligands and maximal unbiased in terms of "artificial enrichment" and "analogue bias". We also conducted comparative studies with DUD-E and DEKOIS 2.0 sets against HDAC2 and HDAC8 targets and demonstrate that our MUBD-HDACs are unique in that they can be applied unbiasedly to both LBVS and SBVS approaches. In addition, we defined a novel metric, i.e. NLBScore, to detect the "2D bias" and "LBVS favorable" effect within the benchmarking sets. In summary, MUBD-HDACs are the only comprehensive and maximal-unbiased benchmark data sets for HDACs (including Sirtuins) that are available so far. MUBD-HDACs are freely available at http://www.xswlab.org/ .

Benchmarking methods and data sets for ligand enrichment assessment in virtual screening

Retrospective small-scale virtual screening (VS) based on benchmarking data sets has been widely used to estimate ligand enrichments of VS approaches in the prospective (i.e. real-world) efforts. However, the intrinsic differences of benchmarking sets to the real screening chemical libraries can cause biased assessment. Herein, we summarize the history of benchmarking methods as well as data sets and highlight three main types of biases found in benchmarking sets, i.e. "analogue bias", "artificial enrichment" and "false negative". In addition, we introduce our recent algorithm to build maximum-unbiased benchmarking sets applicable to both ligand-based and structure-based VS approaches, and its implementations to three important human histone deacetylases (HDACs) isoforms, i.e. HDAC1, HDAC6 and HDAC8. The leave-one-out cross-validation (LOO CV) demonstrates that the benchmarking sets built by our algorithm are maximum-unbiased as measured by property matching, ROC curves and AUCs.

Shaping the interaction landscape of bioactive molecules

MOTIVATION

Most bioactive molecules perform their action by interacting with proteins or other macromolecules. However, for a significant fraction of them, the primary target remains unknown. In addition, the majority of bioactive molecules have more than one target, many of which are poorly characterized. Computational predictions of bioactive molecule targets based on similarity with known ligands are powerful to narrow down the number of potential targets and to rationalize side effects of known molecules.

RESULTS

Using a reference set of 224 412 molecules active on 1700 human proteins, we show that accurate target prediction can be achieved by combining different measures of chemical similarity based on both chemical structure and molecular shape. Our results indicate that the combined approach is especially efficient when no ligand with the same scaffold or from the same chemical series has yet been discovered. We also observe that different combinations of similarity measures are optimal for different molecular properties, such as the number of heavy atoms. This further highlights the importance of considering different classes of similarity measures between new molecules and known ligands to accurately predict their targets.

Surflex-Dock: Docking benchmarks and real-world application

Benchmarks for molecular docking have historically focused on re-docking the cognate ligand of a well-determined protein-ligand complex to measure geometric pose prediction accuracy, and measurement of virtual screening performance has been focused on increasingly large and diverse sets of target protein structures, cognate ligands, and various types of decoy sets. Here, pose prediction is reported on the Astex Diverse set of 85 protein ligand complexes, and virtual screening performance is reported on the DUD set of 40 protein targets. In both cases, prepared structures of targets and ligands were provided by symposium organizers. The re-prepared data sets yielded results not significantly different than previous reports of Surflex-Dock on the two benchmarks. Minor changes to protein coordinates resulting from complex pre-optimization had large effects on observed performance, highlighting the limitations of cognate ligand re-docking for pose prediction assessment. Docking protocols developed for cross-docking, which address protein flexibility and produce discrete families of predicted poses, produced substantially better performance for pose prediction. Performance on virtual screening performance was shown to benefit by employing and combining multiple screening methods: docking, 2D molecular similarity, and 3D molecular similarity. In addition, use of multiple protein conformations significantly improved screening enrichment.

How to benchmark methods for structure-based virtual screening of large compound libraries

Structure-based virtual screening is a useful computational technique for ligand discovery. To systematically evaluate different docking approaches, it is important to have a consistent benchmarking protocol that is both relevant and unbiased. Here, we describe the designing of a benchmarking data set for docking screen assessment, a standard docking screening process, and the analysis and presentation of the enrichment of annotated ligands among a background decoy database.

Does your model weigh the same as a duck?

Computer-aided drug design is a mature field by some measures, and it has produced notable successes that underpin the study of interactions between small molecules and living systems. However, unlike a truly mature field, fallacies of logic lie at the heart of the arguments in support of major lines of research on methodology and validation thereof. Two particularly pernicious ones are cum hoc ergo propter hoc (with this, therefore because of this) and confirmation bias (seeking evidence that is confirmatory of the hypothesis at hand). These fallacies will be discussed in the context of off-target predictive modeling, QSAR, molecular similarity computations, and docking. Examples will be shown that avoid these problems.

Identifying and characterizing promiscuous targets: implications for virtual screening

INTRODUCTION

Ligand-based shape matching approaches have become established as important and popular virtual screening (VS) techniques. However, despite their relative success, the question of how to best choose the initial query compounds and their conformations remains largely unsolved. This issue gains importance when dealing with promiscuous targets, that is, proteins that bind multiple ligand scaffold families in one or more binding site. Conventional shape matching VS approaches assume that there is only one binding mode for a given protein target. This may be true for some targets, but it is certainly not true in all cases. Several recent studies have shown that some protein targets bind to different ligands in different ways.

AREAS COVERED

The authors discuss the concept of promiscuity in the context of virtual drug screening, and present and analyze several examples of promiscuous targets. The article also reports on the impact of the query conformation on the performance of shape-based VS and the potential to improve VS performance by using consensus shape clustering techniques.

EXPERT OPINION

The notion of polypharmacology is becoming highly relevant in drug discovery. Understanding and exploiting promiscuity present challenges and opportunities for drug discovery endeavors. The examples of promiscuity presented here suggest that promiscuous targets and ligands are much more common than previously assumed, and this should be taken into account in practical VS protocols. Although some progress has been made, there is a need to develop more sophisticated computational techniques and protocols that can identify and characterize promiscuous targets on a genomic scale.

Chemical structural novelty: on-targets and off-targets

Drug structures may be quantitatively compared based on 2D topological structural considerations and based on 3D characteristics directly related to binding. A framework for combining multiple similarity computations is presented along with its systematic application to 358 drugs with overlapping pharmacology. Given a new molecule along with a set of molecules sharing some biological effect, a single score based on comparison to the known set is produced, reflecting either 2D similarity, 3D similarity, or their combination. For prediction of primary targets, the benefit of 3D over 2D was relatively small, but for prediction of off-targets, the added benefit was large. In addition to assessing prediction, the relationship between chemical similarity and pharmacological novelty was studied. Drug pairs that shared high 3D similarity but low 2D similarity (i.e., a novel scaffold) were shown to be much more likely to exhibit pharmacologically relevant differences in terms of specific protein target modulation.

Effectiveness of 2D fingerprints for scaffold hopping

Background: It has been suggested that similarity searching using 2D fingerprints may not be suitable for scaffold hopping. Methods: This article reports a detailed evaluation of the effectiveness of six common types of 2D fingerprints when they are used for scaffold-hopping similarity searches of the Molecular Design Limited Drug Data Report database, World of Molecular Bioactivity database and Maximum Unbiased Validation database. Results: The results demonstrate that 2D fingerprints can be used for scaffold hopping, with novel scaffolds being identified in nearly every search that was carried out. The degree of enrichment depends on the structural diversity of the actives that are being sought, with the greatest enrichments often being obtained using the extended connectivity fingerprint encoding a circular substructure of diameter four bonds (ECFP4) fingerprint. Conclusion: 2D fingerprints provide a simple and computationally efficient way of identifying novel chemotypes in lead-discovery programs.

FRED pose prediction and virtual screening accuracy

Results of a previous docking study are reanalyzed and extended to include results from the docking program FRED and a detailed statistical analysis of both structure reproduction and virtual screening results. FRED is run both in a traditional docking mode and in a hybrid mode that makes use of the structure of a bound ligand in addition to the protein structure to screen molecules. This analysis shows that most docking programs are effective overall but highly inconsistent, tending to do well on one system and poorly on the next. Comparing methods, the difference in mean performance on DUD is found to be statistically significant (95% confidence) 61% of the time when using a global enrichment metric (AUC). Early enrichment metrics are found to have relatively poor statistical power, with 0.5% early enrichment only able to distinguish methods to 95% confidence 14% of the time.

Comparative evaluation of 3D virtual ligand screening methods: impact of the molecular alignment on enrichment

In the early stage of drug discovery programs, when the structure of a complex involving a target and a small molecule is available, structure-based virtual ligand screening methods are generally preferred. However, ligand-based strategies like shape-similarity search methods can also be applied. Shape-similarity search methods consist in exploring a pseudo-binding-site derived from the known small molecule used as a reference. Several of these methods use conformational sampling algorithms which are also shared by corresponding docking methods: for example Surflex-dock/Surflex-sim, FlexX/FlexS, ICM, and OMEGA-FRED/OMEGA-ROCS. Using 11 systems issued from the challenging "own" subsets of the Directory of Useful Decoys (DUD-own), we evaluated and compared the performance of the above-cited programs in terms of molecular alignment accuracy, enrichment in active compounds, and enrichment in different chemotypes (scaffold-hopping). Since molecular alignment is a crucial aspect of performance for the different methods, we have assessed its impact on enrichment. We have also illustrated the paradox of retrieving active compounds with good scores even if they are inaccurately positioned. Finally, we have highlighted possible positive aspects of using shape-based approaches in drug-discovery protocols when the structure of the target in complex with a small molecule is known.

Molecular shape and medicinal chemistry: a perspective

The eight contributions here provide ample evidence that shape as a volume or as a surface is a vibrant and useful concept when applied to drug discovery. It provides a reliable scaffold for "decoration" with chemical intuition (or bias) for virtual screening and lead optimization but also has its unadorned uses, as in library design, ligand fitting, pose prediction, or active site description. Computing power has facilitated this evolution by allowing shape to be handled precisely without the need to reduce down to point descriptors or approximate metrics, and the diversity of resultant applications argues for this being an important step forward. Certainly, it is encouraging that as computation has enabled our intuition, molecular shape has consistently surprised us in its usefulness and adaptability. The first Aurelius question, "What is the essence of a thing?", seems well answered, however, the third, "What do molecules do?", only partly so. Are the topics covered here exhaustive, or is there more to come? To date, there has been little published on the use of the volumetric definition of shape described here as a QSAR variable, for instance, in the prediction or classification of activity, although other shape definitions have been successful applied, for instance, as embodied in the Compass program described above in "Shape from Surfaces". Crystal packing is a phenomenon much desired to be understood. Although powerful models have been applied to the problem, to what degree is this dominated purely by the shape of a molecule? The shape comparison described here is typically of a global nature, and yet some importance must surely be placed on partial shape matching, just as the substructure matching of chemical graphs has proved useful. The approach of using surfaces, as described here, offers some flavor of this, as does the use of metrics that penalize volume mismatch less than the Tanimoto, e.g., Tversky measures. As yet, there is little to go on as to how useful a paradigm this will be because there is less software and fewer concrete results.Finally, the distance between molecular shapes, or between any shapes defined as volumes or surfaces, is a metric property in the mathematical sense of the word. As yet, there has been little, if any, application of this observation. We cannot know what new application to the design and discovery of pharmaceuticals may yet arise from the simple concept of molecular shape, but it is fair to say that the progress so far is impressive.

Physical binding pocket induction for affinity prediction

Computational methods for predicting ligand affinity where no protein structure is known generally take the form of regression analysis based on molecular features that have only a tangential relationship to a protein/ligand binding event. Such methods have limited utility when structural variation moves beyond congeneric series. We present a novel approach based on the multiple-instance learning method of Compass, where a physical model of a binding site is induced from ligands and their corresponding activity data. The model consists of molecular fragments that can account for multiple positions of literal protein residues. We demonstrate the method on 5HT1a ligands by training on a series with limited scaffold variation and testing on numerous ligands with variant scaffolds. Predictive error was between 0.5 and 1.0 log units (0.7-1.4 kcal/mol), with statistically significant rank correlations. Accurate activity predictions of novel ligands were demonstrated using a validation approach where a small number of ligands of limited structural variation known at a fixed time point were used to make predictions on a blind test set of widely varying molecules, some discovered at a much later time point.

Optimal assignment methods for ligand-based virtual screening

BACKGROUND

Ligand-based virtual screening experiments are an important task in the early drug discovery stage. An ambitious aim in each experiment is to disclose active structures based on new scaffolds. To perform these "scaffold-hoppings" for individual problems and targets, a plethora of different similarity methods based on diverse techniques were published in the last years. The optimal assignment approach on molecular graphs, a successful method in the field of quantitative structure-activity relationships, has not been tested as a ligand-based virtual screening method so far.

RESULTS

We evaluated two already published and two new optimal assignment methods on various data sets. To emphasize the "scaffold-hopping" ability, we used the information of chemotype clustering analyses in our evaluation metrics. Comparisons with literature results show an improved early recognition performance and comparable results over the complete data set. A new method based on two different assignment steps shows an increased "scaffold-hopping" behavior together with a good early recognition performance.

CONCLUSION

The presented methods show a good combination of chemotype discovery and enrichment of active structures. Additionally, the optimal assignment on molecular graphs has the advantage to investigate and interpret the mappings, allowing precise modifications of internal parameters of the similarity measure for specific targets. All methods have low computation times which make them applicable to screen large data sets.

Discovery of novel triple helical DNA intercalators by an integrated virtual and actual screening platform

Virtual Screening is an increasingly attractive way to discover new small molecules with potential medicinal value. We introduce a novel strategy that integrates use of the molecular docking software Surflex with experimental validation by the method of competition dialysis. This integrated approach was used to identify ligands that selectively bind to the triplex DNA poly(dA)-[poly(dT)]₂. A library containing ∼2 million ligands was virtually screened to identify compounds with chemical and structural similarity to a known triplex intercalator, the napthylquinoline MHQ-12. Further molecular docking studies using compounds with high structural similarity resulted in two compounds that were then demonstrated by competition dialysis to have a superior affinity and selectivity for the triplex nucleic acid than MHQ-12. One of the compounds has a different chemical backbone than MHQ-12, which demonstrates the ability of this strategy to ‘scaffold hop’ and to identify small molecules with novel binding properties. Biophysical characterization of these compounds by circular dichroism and thermal denaturation studies confirmed their binding mode and selectivity. These studies provide a proof-of-principle for our integrated screening strategy, and suggest that this platform may be extended to discover new compounds that target therapeutically relevant nucleic acid morphologies.

Selective structure-based virtual screening for full and partial agonists of the beta2 adrenergic receptor

The recently solved high-resolution X-ray structure of the beta2 adrenergic receptor has been challenged for its ability to discriminate inverse agonists/antagonists from partial/full agonists. Whereas the X-ray structure of the ground state receptor was unsuitable to distinguish true ligands with different functional effects, modifying this structure to reflect early conformational events in receptor activation led to a receptor model able to selectively retrieve full and partial agonists by structure-based virtual screening. The use of a topological scoring function based on molecular interaction fingerprints was shown to be mandatory to properly rank docking poses and achieve acceptable enrichments for partial and full agonists only.