IChem: A Versatile Toolkit for Detecting, Comparing, and Predicting Protein-Ligand Interactions

InDeepNet: a web platform for predicting functional binding sites in proteins using InDeep

Predicting functional binding sites in proteins is crucial for understanding protein-protein interactions (PPIs) and identifying drug targets. While various computational approaches exist, many fail to assess PPI ligandability, which often involves conformational changes. We introduce InDeepNet, a web-based platform integrating InDeep, a deep-learning model for binding site prediction, with InDeepHolo, which evaluates a site's propensity to adopt a ligand-bound (holo) conformation. InDeepNet provides an intuitive interface for researchers to upload protein structures from in-house data, the Protein Data Bank (PDB), or AlphaFold, predicting potential binding sites for proteins or small molecules. Results are presented as interactive 3D visualizations via Mol*, facilitating structural analysis. With InDeepHolo, the platform helps select conformations optimal for small-molecule binding, improving structure-based drug design. Accessible at https://indeep-net.gpu.pasteur.cloud/, InDeepNet removes the need for specialized coding skills or high-performance computing, making advanced predictive models widely available. By streamlining PPI target assessment and ligandability prediction, it assists research and supports therapeutic development targeting PPIs.

GRADE and X-GRADE: Unveiling Novel Protein-Ligand Interaction Fingerprints Based on GRAIL Scores

Nonbonding molecular interactions, such as hydrogen bonding, hydrophobic contacts, ionic interactions, etc., are at the heart of many biological processes, and their appropriate treatment is essential for the successful application of numerous computational drug design methods. This paper introduces GRADE, a novel interaction fingerprint (IFP) descriptor that quantifies these interactions using floating point values derived from GRAIL scores, encoding both the presence and quality of interactions. GRADE is available in two versions: a basic 35-element variant and an extended 177-element variant. Three case studies demonstrate GRADE's utility: (1) dimensionality reduction for visualizing the chemical space of protein-ligand complexes using Uniform Manifold Approximation and Projection (UMAP), showing competitive performance with complex descriptors; (2) binding affinity prediction, where GRADE achieved reasonable accuracy with minimal machine learning optimization; and (3) three-dimensional-quantitative structure-activity relationship (3D-QSAR) modeling for a specific protein target, where GRADE enhanced the performance of Morgan Fingerprints.

Structural basis of positive allosteric modulation of metabotropic glutamate receptor activation and internalization

The metabotropic glutamate receptors (mGluRs) are neuromodulatory family C G protein coupled receptors which assemble as dimers and allosterically couple extracellular ligand binding domains (LBDs) to transmembrane domains (TMDs) to drive intracellular signaling. Pharmacologically, mGluRs can be targeted at the LBDs by glutamate and synthetic orthosteric compounds or at the TMDs by allosteric modulators. Despite the potential of allosteric compounds as therapeutics, an understanding of the functional and structural basis of their effects is limited. Here we use multiple approaches to dissect the functional and structural effects of orthosteric versus allosteric ligands. We find, using electrophysiological and live cell imaging assays, that both agonists and positive allosteric modulators (PAMs) can drive activation and internalization of group II and III mGluRs. The effects of PAMs are pleiotropic, boosting the maximal response to orthosteric agonists and serving independently as internalization-biased agonists across mGluR subtypes. Motivated by this and intersubunit FRET analyses, we determine cryo-electron microscopy structures of mGluR3 in the presence of either an agonist or antagonist alone or in combination with a PAM. These structures reveal PAM-driven re-shaping of intra- and inter-subunit conformations and provide evidence for a rolling TMD dimer interface activation pathway that controls G protein and beta-arrestin coupling.

Subpocket Similarity-Based Hit Identification for Challenging Targets: Application to the WDR Domain of LRRK2

We herewith applied a priori a generic hit identification method (POEM) for difficult targets of known three-dimensional structure, relying on the simple knowledge of physicochemical and topological properties of a user-selected cavity. Searching for local similarity to a set of fragment-bound protein microenvironments of known structure, a point cloud registration algorithm is first applied to align known subpockets to the target cavity. The resulting alignment then permits us to directly pose the corresponding seed fragments in a target cavity space not typically amenable to classical docking approaches. Last, linking potentially connectable atoms by a deep generative linker enables full ligand enumeration. When applied to the WD40 repeat (WDR) central cavity of leucine-rich repeat kinase 2 (LRRK2), an unprecedented binding site, POEM was able to quickly propose 94 potential hits, five of which were subsequently confirmed to bind in vitro to LRRK2-WDR.

LVPocket: integrated 3D global-local information to protein binding pockets prediction with transfer learning of protein structure classification

BACKGROUND

Previous deep learning methods for predicting protein binding pockets mainly employed 3D convolution, yet an abundance of convolution operations may lead the model to excessively prioritize local information, thus overlooking global information. Moreover, it is essential for us to account for the influence of diverse protein folding structural classes. Because proteins classified differently structurally exhibit varying biological functions, whereas those within the same structural class share similar functional attributes.

RESULTS

We proposed LVPocket, a novel method that synergistically captures both local and global information of protein structure through the integration of Transformer encoders, which help the model achieve better performance in binding pockets prediction. And then we tailored prediction models for data of four distinct structural classes of proteins using the transfer learning. The four fine-tuned models were trained on the baseline LVPocket model which was trained on the sc-PDB dataset. LVPocket exhibits superior performance on three independent datasets compared to current state-of-the-art methods. Additionally, the fine-tuned model outperforms the baseline model in terms of performance.

SCIENTIFIC CONTRIBUTION

We present a novel model structure for predicting protein binding pockets that provides a solution for relying on extensive convolutional computation while neglecting global information about protein structures. Furthermore, we tackle the impact of different protein folding structures on binding pocket prediction tasks through the application of transfer learning methods.

Structural analysis of neomycin B and kanamycin A binding Aminoglycosides Modifying Enzymes (AME) and bacterial ribosomal RNA

Aminoglycosides are crucial antibiotics facing challenges from bacterial resistance. This study addresses the importance of aminoglycoside modifying enzymes in the context of escalating resistance. Drawing upon over two decades of structural data in the Protein Data Bank, we focused on two key antibiotics, neomycin B and kanamycin A, to explore how the aminoglycoside structure is exploited by this family of enzymes. A systematic comparison across diverse enzymes and the RNA A-site target identified common characteristics in the recognition mode, while assessing the adaptability of neomycin B and kanamycin A in various environments.

Protein Structure-Based Organic Chemistry-Driven Ligand Design from Ultralarge Chemical Spaces

Ultralarge chemical spaces describing several billion compounds are revolutionizing hit identification in early drug discovery. Because of their size, such chemical spaces cannot be fully enumerated and require ad-hoc computational tools to navigate them and pick potentially interesting hits. We here propose a structure-based approach to ultralarge chemical space screening in which commercial chemical reagents are first docked to the target of interest and then directly connected according to organic chemistry and topological rules, to enumerate drug-like compounds under three-dimensional constraints of the target. When applied to bespoke chemical spaces of different sizes and chemical complexity targeting two receptors of pharmaceutical interest (estrogen β receptor, dopamine D3 receptor), the computational method was able to quickly enumerate hits that were either known ligands (or very close analogs) of targeted receptors as well as chemically novel candidates that could be experimentally confirmed by in vitro binding assays. The proposed approach is generic, can be applied to any docking algorithm, and requires few computational resources to prioritize easily synthesizable hits from billion-sized chemical spaces.

Assessing the role of residue Phe108 of cytochrome P450 3A4 in allosteric effects of midazolam metabolism

Cytochrome P450 3A4 (CYP3A4) is involved in the metabolism of more drugs in clinical use than any other xenobiotic-metabolizing enzyme. CYP3A4-mediated drug metabolism is usually allosterically modulated by substrate concentration (homotropic allostery) and other drugs (heterotropic allostery), exhibiting unusual kinetic profiles and regiospecific metabolism. Recent studies suggest that residue Phe108 (F108) of CYP3A4 may have an important role in drug metabolism. In this work, residue mutations were coupled with well-tempered metadynamics simulations to assess the importance of F108 in the allosteric effects of midazolam metabolism. Comparing the simulation results of the wild-type and mutation systems, we identify that the π-π interaction and steric effect between the F108 side chain and midazolam is favorable for the stable binding of substrate in the active site. F108 also plays an important role in the transition of substrate binding mode, which mainly induces the transition of substrate binding mode by forming π-π interactions with multiple aromatic rings of the substrate. Moreover, the side chain of F108 is closely related to the radius and depth of the 2a and 2f channels, and F108 may further regulate drug metabolism by affecting the pathway, orientation, or time of substrate entry into the CYP3A4 active site or product egress from the active site. Altogether, we suggest that F108 affects drug metabolism and regulatory mechanisms by affecting substrate binding stability, binding mode transition, and channel characteristics of CYP3A4. Our findings could promote the understanding of complicated allosteric mechanisms in CYP3A4-mediated drug metabolism, and the knowledge could be used for drug development and disease treatment.

Fingerprinting Interactions between Proteins and Ligands for Facilitating Machine Learning in Drug Discovery

Molecular recognition is fundamental in biology, underpinning intricate processes through specific protein-ligand interactions. This understanding is pivotal in drug discovery, yet traditional experimental methods face limitations in exploring the vast chemical space. Computational approaches, notably quantitative structure-activity/property relationship analysis, have gained prominence. Molecular fingerprints encode molecular structures and serve as property profiles, which are essential in drug discovery. While two-dimensional (2D) fingerprints are commonly used, three-dimensional (3D) structural interaction fingerprints offer enhanced structural features specific to target proteins. Machine learning models trained on interaction fingerprints enable precise binding prediction. Recent focus has shifted to structure-based predictive modeling, with machine-learning scoring functions excelling due to feature engineering guided by key interactions. Notably, 3D interaction fingerprints are gaining ground due to their robustness. Various structural interaction fingerprints have been developed and used in drug discovery, each with unique capabilities. This review recapitulates the developed structural interaction fingerprints and provides two case studies to illustrate the power of interaction fingerprint-driven machine learning. The first elucidates structure-activity relationships in β2 adrenoceptor ligands, demonstrating the ability to differentiate agonists and antagonists. The second employs a retrosynthesis-based pre-trained molecular representation to predict protein-ligand dissociation rates, offering insights into binding kinetics. Despite remarkable progress, challenges persist in interpreting complex machine learning models built on 3D fingerprints, emphasizing the need for strategies to make predictions interpretable. Binding site plasticity and induced fit effects pose additional complexities. Interaction fingerprints are promising but require continued research to harness their full potential.

Predicting the duration of action of β2-adrenergic receptor agonists: Ligand and structure-based approaches

Agonists of the β2 adrenergic receptor (ADRB2) are an important class of medications used for the treatment of respiratory diseases. They can be classified as short acting (SABA) or long acting (LABA), with each class playing a different role in patient management. In this work we explored both ligand-based and structure-based high-throughput approaches to classify β2-agonists based on their duration of action. A completely in-silico prediction pipeline using an AlphaFold generated structure was used for structure-based modelling. Our analysis identified the ligands' 3D structure and lipophilicity as the most relevant features for the prediction of the duration of action. Interaction-based methods were also able to select ligands with the desired duration of action, incorporating the bias directly in the structure-based drug discovery pipeline without the need for further processing.

The Impact of Data on Structure-Based Binding Affinity Predictions Using Deep Neural Networks

Artificial intelligence (AI) has gained significant traction in the field of drug discovery, with deep learning (DL) algorithms playing a crucial role in predicting protein-ligand binding affinities. Despite advancements in neural network architectures, system representation, and training techniques, the performance of DL affinity prediction has reached a plateau, prompting the question of whether it is truly solved or if the current performance is overly optimistic and reliant on biased, easily predictable data. Like other DL-related problems, this issue seems to stem from the training and test sets used when building the models. In this work, we investigate the impact of several parameters related to the input data on the performance of neural network affinity prediction models. Notably, we identify the size of the binding pocket as a critical factor influencing the performance of our statistical models; furthermore, it is more important to train a model with as much data as possible than to restrict the training to only high-quality datasets. Finally, we also confirm the bias in the typically used current test sets. Therefore, several types of evaluation and benchmarking are required to understand models' decision-making processes and accurately compare the performance of models.

How Ligands Interact with the Kinase Hinge

ATP-competitive kinase inhibitors form hydrogen bond interactions with the kinase hinge region at the adenine binding site. Thus, it is crucial to explore hinge-ligand recognition as part of a rational drug design strategy. Here, harnessing known ligand-bound kinase structures and experimental assay resources, we first created a kinase structure-assay database (KSAD) containing 2705 nM ligand-bound kinase complexes. Then, using KSAD, we systematically investigate hinge-ligand binding patterns using interaction fingerprints, thereby delineating 15 different hydrogen-bond interaction modes. We believe these results will be valuable for de novo drug design and/or scaffold hopping of kinase-targeted drugs.

A practical guide to machine-learning scoring for structure-based virtual screening

Structure-based virtual screening (SBVS) via docking has been used to discover active molecules for a range of therapeutic targets. Chemical and protein data sets that contain integrated bioactivity information have increased both in number and in size. Artificial intelligence and, more concretely, its machine-learning (ML) branch, including deep learning, have effectively exploited these data sets to build scoring functions (SFs) for SBVS against targets with an atomic-resolution 3D model (e.g., generated by X-ray crystallography or predicted by AlphaFold2). Often outperforming their generic and non-ML counterparts, target-specific ML-based SFs represent the state of the art for SBVS. Here, we present a comprehensive and user-friendly protocol to build and rigorously evaluate these new SFs for SBVS. This protocol is organized into four sections: (i) using a public benchmark of a given target to evaluate an existing generic SF; (ii) preparing experimental data for a target from public repositories; (iii) partitioning data into a training set and a test set for subsequent target-specific ML modeling; and (iv) generating and evaluating target-specific ML SFs by using the prepared training-test partitions. All necessary code and input/output data related to three example targets (acetylcholinesterase, HMG-CoA reductase, and peroxisome proliferator-activated receptor-α) are available at https://github.com/vktrannguyen/MLSF-protocol , can be run by using a single computer within 1 week and make use of easily accessible software/programs (e.g., Smina, CNN-Score, RF-Score-VS and DeepCoy) and web resources. Our aim is to provide practical guidance on how to augment training data to enhance SBVS performance, how to identify the most suitable supervised learning algorithm for a data set, and how to build an SF with the highest likelihood of discovering target-active molecules within a given compound library.

De Novo Design of κ-Opioid Receptor Antagonists Using a Generative Deep-Learning Framework

Likely effective pharmacological interventions for the treatment of opioid addiction include attempts to attenuate brain reward deficits during periods of abstinence. Pharmacological blockade of the κ-opioid receptor (KOR) has been shown to abolish brain reward deficits in rodents during withdrawal, as well as to reduce the escalation of opioid use in rats with extended access to opioids. Although KOR antagonists represent promising candidates for the treatment of opioid addiction, very few potent selective KOR antagonists are known to date and most of them exhibit significant safety concerns. Here, we used a generative deep-learning framework for the de novo design of chemotypes with putative KOR antagonistic activity. Molecules generated by models trained with this framework were prioritized for chemical synthesis based on their predicted optimal interactions with the receptor. Our models and proposed training protocol were experimentally validated by binding and functional assays.

Structural basis of allosteric modulation of metabotropic glutamate receptor activation and desensitization

The metabotropic glutamate receptors (mGluRs) are neuromodulatory family C G protein coupled receptors which assemble as dimers and allosterically couple extracellular ligand binding domains (LBDs) to transmembrane domains (TMDs) to drive intracellular signaling. Pharmacologically, mGluRs can be targeted either at the LBDs by glutamate and synthetic orthosteric compounds or at the TMDs by allosteric modulators. Despite the potential of allosteric TMD-targeting compounds as therapeutics, an understanding of the functional and structural basis of their effects on mGluRs is limited. Here we use a battery of approaches to dissect the distinct functional and structural effects of orthosteric versus allosteric ligands. We find using electrophysiological and live cell imaging assays that both agonists and positive allosteric modulators (PAMs) can drive activation and desensitization of mGluRs. The effects of PAMs are pleiotropic, including both the ability to boost the maximal response to orthosteric agonists and to serve independently as desensitization-biased agonists across mGluR subtypes. Conformational sensors reveal PAM-driven inter-subunit re-arrangements at both the LBD and TMD. Motivated by this, we determine cryo-electron microscopy structures of mGluR3 in the presence of either an agonist or antagonist alone or in combination with a PAM. These structures reveal PAM-driven re-shaping of intra- and inter-subunit conformations and provide evidence for a rolling TMD dimer interface activation pathway that controls G protein and beta-arrestin coupling.

Highlights

-Agonists and PAMs drive mGluR activation, desensitization, and endocytosis-PAMs are desensitization-biased and synergistic with agonists-Four combinatorial ligand conditions reveal an ensemble of full-length mGluR structures with novel interfaces-Activation and desensitization involve rolling TMD interfaces which are re-shaped by PAM.

De Novo Design of κ-Opioid Receptor Antagonists Using a Generative Deep Learning Framework

Likely effective pharmacological interventions for the treatment of opioid addiction include attempts to attenuate brain reward deficits during periods of abstinence. Pharmacological blockade of the κ-opioid receptor (KOR) has been shown to abolish brain reward deficits in rodents during withdrawal, as well as to reduce the escalation of opioid use in rats with extended access to opioids. Although KOR antagonists represent promising candidates for the treatment of opioid addiction, very few potent selective KOR antagonists are known to date and most of them exhibit significant safety concerns. Here, we used a generative deep learning framework for the de novo design of chemotypes with putative KOR antagonistic activity. Molecules generated by models trained with this framework were prioritized for chemical synthesis based on their predicted optimal interactions with the receptor. Our models and proposed training protocol were experimentally validated by binding and functional assays.

Analysis of KRAS-Ligand Interaction Modes and Flexibilities Reveals the Binding Characteristics

KRAS, a common human oncogene, has been recognized as a critical drug target in treating multiple cancers. After four decades of effort, one allosteric KRAS drug (Sotorasib) has been approved, inspiring more KRAS-targeted drug research. Here, we provide the features of KRAS binding pockets and ligand-binding characteristics of KRAS complexes using a structural systems pharmacology approach. Three distinct binding sites (conserved nucleotide-binding site, shallow Switch-I/II pocket, and allosteric Switch-II/α3 pocket) are characterized. Ligand-binding features are determined based on encoded KRAS-inhibitor interaction fingerprints. Finally, the flexibility of the three distinct binding sites to accommodate different potential ligands, based on MD simulation, is discussed. Collectively, these findings are intended to facilitate rational KRAS drug design.

Mining the Protein Data Bank to inspire fragment library design

The fragment approach has emerged as a method of choice for drug design, as it allows difficult therapeutic targets to be addressed. Success lies in the choice of the screened chemical library and the biophysical screening method, and also in the quality of the selected fragment and structural information used to develop a drug-like ligand. It has recently been proposed that promiscuous compounds, i.e., those that bind to several proteins, present an advantage for the fragment approach because they are likely to give frequent hits in screening. In this study, we searched the Protein Data Bank for fragments with multiple binding modes and targeting different sites. We identified 203 fragments represented by 90 scaffolds, some of which are not or hardly present in commercial fragment libraries. By contrast to other available fragment libraries, the studied set is enriched in fragments with a marked three-dimensional character (download at 10.5281/zenodo.7554649).

Exploring cooperative molecular contacts using a PostgreSQL database system

Cooperative molecular contacts play an important role in protein structure and ligand binding. Here, we constructed a PostgreSQL database that stores structural information in the form of atomic environments and allows flexible mining of molecular contacts. Taking the Ser-His-Asp/Glu catalytic triad as a first test case, we demonstrate that the presence of a carboxylate oxygen atom in the vicinity of a His is associated with shorter Ser-OH..N-His bond in the PDB30 subset. We prospectively mine catalytic triads in unannotated proteins, suggesting catalytic functions for unannotated proteins. As a second test case, we demonstrate that this database system can include ligand atoms, represented by Sybyl atom types, by evaluating the proportion of counter-ions for ligand carboxylate oxygens.

Discovery of potent microtubule-destabilizing agents targeting for colchicine site by virtual screening, biological evaluation, and molecular dynamics simulation

Microtubule has been considered as attractive therapeutic target for various cancers. Although numerous of chemically diverse compounds targeting to colchicine site have been reported, none of them was approved by Food and Drug Administration. In this investigation, the virtual screening methods, including pharmacophore model, molecular docking, and interaction molecular fingerprints similarity, were applied to discover novel microtubule-destabilizing agents from database with 324,474 compounds. 22 compounds with novel scaffolds were identified as microtubule-destabilizing agents, and then submitted to the biological evaluation. Among these 22 hits, hit4 with novel scaffold represents the best anti-proliferative activity with IC₅₀ ranging from 4.51 to 14.81 μM on four cancer cell lines. The in vitro assays reveal that hit4 can effectively inhibit tubulin assembly, and disrupt the microtubule network in MCF-7 cell at a concentration-dependent manner. Finally, the molecular dynamics simulation analysis exhibits that hit4 can stably bind to colchicine site, interact with key residues, and induce αT5 and βT7 regions changes. The values of ΔG_bind for the tubulin-colchicine and tubulin-hit4 are -172.9±10.5 and -166.0±12.6 kJ·mol^-1, respectively. The above results indicate that the hit4 is a novel microtubule destabilizing agent targeting to colchicine-binding site, which could be developed as a promising tubulin polymerization inhibitor with higher activity for cancer therapy.

SAR exploration of the non-imidazole histamine H3 receptor ligand ZEL-H16 reveals potent inverse agonism

Histamine H₃ receptor (H₃ R) agonists without an imidazole moiety remain very scarce. Of these, ZEL-H16 (1) has been reported previously as a high-affinity non-imidazole H₃ R (partial) agonist. Our structure-activity relationship analysis using derivatives of 1 identified both basic moieties as key interaction motifs and the distance of these from the central core as a determinant for H₃ R affinity. However, in spite of the reported H₃ R (partial) agonism, in our hands, 1 acts as an inverse agonist for Gα_i signaling in a CRE-luciferase reporter gene assay and using an H₃ R conformational sensor. Inverse agonism was also observed for all of the synthesized derivatives of 1. Docking studies and molecular dynamics simulations suggest ionic interactions/hydrogen bonds to H₃ R residues D114^3.32 and E206^5.46 as essential interaction points.

Estimating the Similarity between Protein Pockets

With the exponential increase in publicly available protein structures, the comparison of protein binding sites naturally emerged as a scientific topic to explain observations or generate hypotheses for ligand design, notably to predict ligand selectivity for on- and off-targets, explain polypharmacology, and design target-focused libraries. The current review summarizes the state-of-the-art computational methods applied to pocket detection and comparison as well as structural druggability estimates. The major strengths and weaknesses of current pocket descriptors, alignment methods, and similarity search algorithms are presented. Lastly, an exhaustive survey of both retrospective and prospective applications in diverse medicinal chemistry scenarios illustrates the capability of the existing methods and the hurdle that still needs to be overcome for more accurate predictions.

Target-Focused Library Design by Pocket-Applied Computer Vision and Fragment Deep Generative Linking

We here describe a computational approach (POEM: Pocket Oriented Elaboration of Molecules) to drive the generation of target-focused libraries while taking advantage of all publicly available structural information on protein-ligand complexes. A collection of 31 384 PDB-derived images with key shapes and pharmacophoric properties, describing fragment-bound microenvironments, is first aligned to the query target cavity by a computer vision method. The fragments of the most similar PDB subpockets are then directly positioned in the query cavity using the corresponding image transformation matrices. Lastly, suitable connectable atoms of oriented fragment pairs are linked by a deep generative model to yield fully connected molecules. POEM was applied to generate a library of 1.5 million potential cyclin-dependent kinase 8 inhibitors. By synthesizing and testing as few as 43 compounds, a few nanomolar inhibitors were quickly obtained with limited resources in just two iterative cycles.

Random-forest model for drug-target interaction prediction via Kullbeck-Leibler divergence

Virtual screening has significantly improved the success rate of early stage drug discovery. Recent virtual screening methods have improved owing to advances in machine learning and chemical information. Among these advances, the creative extraction of drug features is important for predicting drug–target interaction (DTI), which is a large-scale virtual screening of known drugs. Herein, we report Kullbeck–Leibler divergence (KLD) as a DTI feature and the feature-driven classification model applicable to DTI prediction. For the purpose, E3FP three-dimensional (3D) molecular fingerprints of drugs as a molecular representation allow the computation of 3D similarities between ligands within each target (Q–Q matrix) to identify the uniqueness of pharmacological targets and those between a query and a ligand (Q–L vector) in DTIs. The 3D similarity matrices are transformed into probability density functions via kernel density estimation as a nonparametric estimation. Each density model can exploit the characteristics of each pharmacological target and measure the quasi-distance between the ligands. Furthermore, we developed a random forest model from the KLD feature vectors to successfully predict DTIs for representative 17 targets (mean accuracy: 0.882, out-of-bag score estimate: 0.876, ROC AUC: 0.990). The method is applicable for 2D chemical similarity.

Harnessing systematic protein-ligand interaction fingerprints for drug discovery

Determining protein-ligand interaction characteristics and mechanisms is crucial to the drug discovery process. Here, we review recent progress and successful applications of a systematic protein-ligand interaction fingerprint (IFP) approach for investigating proteome-wide protein-ligand interactions for drug development. Specifically, we review the use of this IFP approach for revealing polypharmacology across the kinome, predicting promising targets from which to design allosteric inhibitors and covalent kinase inhibitors, uncovering the binding mechanisms of drugs of interest, and demonstrating resistant mechanisms of specific drugs. Together, we demonstrate that the IFP strategy is efficient and practical for drug design research for protein kinases as targets and is extensible to other protein families.

Molecular Docking, Molecular Dynamics Simulations, and Free Energy Calculation Insights into the Binding Mechanism between VS-4718 and Focal Adhesion Kinase

Focal adhesion kinase (FAK) is a 125 kDa nonreceptor tyrosine kinase that plays an important role in many carcinomas. Thus, the targeting of FAK by small molecules is considered to be promising for cancer therapy. Some FAK inhibitors have been reported as potential anticancer drugs and have entered into clinical development; for example, VS-4718 is currently undergoing clinical trials. However, the lack of crystal structural data for the binding of VS-4718 with FAK has hindered the optimization of this anticancer agent. In this work, the VS-4718/FAK interaction model was obtained by molecular docking and molecular dynamics simulations. The binding free energies of VS-4718/FAK were also calculated using the molecular mechanics generalized Born surface area method. It was found that the aminopyrimidine group formed hydrogen bonds with the C502 residue of the hinge loop, while the D564 residue of the T-loop interacted with the amide group. In addition, I428, A452, V484, M499, G505, and L553 residues formed hydrophobic interactions with VS-4718. The obtained results therefore provide an improved understanding of the interaction between human FAK and VS-4718. Based on the obtained binding mechanism, 47 novel compounds were designed to target the adenosine 5'-triphosphate-binding pocket of human FAK, and ensemble docking was performed to assess the effects of these modifications on the inhibitor binding affinity. This work is also expected to provide additional insights into potential future target design strategies based on VS-4718.

On the Frustration to Predict Binding Affinities from Protein-Ligand Structures with Deep Neural Networks

Accurate prediction of binding affinities from protein-ligand atomic coordinates remains a major challenge in early stages of drug discovery. Using modular message passing graph neural networks describing both the ligand and the protein in their free and bound states, we unambiguously evidence that an explicit description of protein-ligand noncovalent interactions does not provide any advantage with respect to ligand or protein descriptors. Simple models, inferring binding affinities of test samples from that of the closest ligands or proteins in the training set, already exhibit good performances, suggesting that memorization largely dominates true learning in the deep neural networks. The current study suggests considering only noncovalent interactions while omitting their protein and ligand atomic environments. Removing all hidden biases probably requires much denser protein-ligand training matrices and a coordinated effort of the drug design community to solve the necessary protein-ligand structures.

Investigation of the tight binding mechanism of a new anticoagulant DD217 to factor Xa by means of molecular docking and molecular dynamics

A new promising drug candidate DD217 has been proposed recently as a potent anticoagulant acting on factor Xa (fXa) target. It exhibits the lowest concentration of doubling the prothrombin time among the known anticoagulants. In order to explain the efficacy of DD217 in terms of molecular interactions with its target we studied the hypothesis of the tight binding mechanism by means of molecular dynamics simulations and statistical analysis of the trajectory. The conducted analysis confirms the significant contributions to the MM/GBSA estimated binding free energy of the S4 pocket residues as well the crucial role of establishing the hydrogen bonds between the ligand and the backbone amides of Gly216 and Gly218 of the target. The simulation results support the hypothesis of the tight binding mechanism of DD217 to fXa.Communicated by Ramaswamy H. Sarma.

BiRDS - Binding Residue Detection from Protein Sequences Using Deep ResNets

Protein-drug interactions play important roles in many biological processes and therapeutics. Predicting the binding sites of a protein helps to discover such interactions. New drugs can be designed to optimize these interactions, improving protein function. The tertiary structure of a protein decides the binding sites available to the drug molecule, but the determination of the 3D structure is slow and expensive. Conversely, the determination of the amino acid sequence is swift and economical. Although quick and accurate prediction of the binding site using just the sequence is challenging, the application of Deep Learning, which has been hugely successful in several biochemical tasks, makes it feasible. BiRDS is a Residual Neural Network that predicts the protein's most active binding site using sequence information. SC-PDB, an annotated database of druggable binding sites, is used for training the network. Multiple Sequence Alignments of the proteins in the database are generated using DeepMSA, and features such as Position-Specific Scoring Matrix, Secondary Structure, and Relative Solvent Accessibility are extracted. During training, a weighted binary cross-entropy loss function is used to counter the substantial imbalance in the two classes of binding and nonbinding residues. A novel test set SC6K is introduced to compare binding-site prediction methods. BiRDS achieves an AUROC score of 0.87, and the center of 25% of its predicted binding sites lie within 4 Å of the center of the actual binding site.

Conformations and binding pockets of HRas and its guanine nucleotide exchange factors complexes in the guanosine triphosphate exchange process

The human Son of Sevenless (SOS) activates the signal-transduction protein Ras by forming the complex SOS·Ras and accelerating the guanosine triphosphate (GTP) exchange in Ras. Inhibition of SOS·Ras could regulate the function of Ras in cells and has emerged as an effective strategy for battling Ras related cancers. A key factor to the success of this approach is to understand the conformational change of Ras during the GTP exchange process. In this study, we perform an extensive molecular dynamics simulation to characterize the specific conformations of Ras without and with guanine nucleotide exchange factors (GEFs) of SOS, especially for the substates of State 1 of HRasGTP∙Mg²⁺ . The potent binding pockets on the surfaces of the RasGDP∙Mg²⁺ , the S1.1 and S1.2 substates in State 1 of RasGTP∙Mg²⁺ and the ternary complexes with SOS are predicted, including the binding sites of other domains of SOS. These findings help to obtain a more thorough understanding of Ras functions in the GTP cycling process and provide a structural foundation for future drug design.

Discovery of microtubule stabilizers with novel scaffold structures based on virtual screening, biological evaluation, and molecular dynamics simulation

Disrupting the dynamics and structures of microtubules can perturb mitotic spindle formation, cause cell cycle arrest in G2/M phase, and subsequently lead to cellular death via apoptosis. In this investigation, the structure-based virtual screening methods, including molecular docking and rescoring, and similarity analysis of interaction molecular fingerprints, were developed to discover novel tubulin inhibitors from ChemDiv database with 1,601,806 compounds. The screened compounds were further filtered by PAINS, ADME/T, Toxscore, SAscore, and Drug-likeness analysis. Finally, 17 hit compounds were selected, and then submitted to the biologic evaluation. Among these hits, the P2 exhibited the strongest antiproliferative activity against four tumor cells including HeLa, HepG2, MCF-7, and A549. The in vitro tubulin polymerization assay revealed P2 could promote tubulin polymerization in a dose dependent manner. Finally, in order to analyze the interaction modes of complexes, the molecular dynamics simulation was performed to investigate the interactions between P2 and tubulin. The molecular dynamics simulation analysis showed that P2 could stably bind to taxane site, induced H6-H7, B9-B10, and M-loop regions changes. The ΔG_bind energies of tubulin-P2 and tubulin-paclitaxel were -68.25 ± 12.98 and -146.05 ± 16.17 kJ mol^-1, respectively, which were in line with the results of the experimental test. Therefore, P2 has been well characterized as lead compounds for developing new tubulin inhibitors with potential anticancer activity.

Structure-based protein-ligand interaction fingerprints for binding affinity prediction

Binding affinity prediction (BAP) using protein–ligand complex structures is crucial to computer-aided drug design, but remains a challenging problem. To achieve efficient and accurate BAP, machine-learning scoring functions (SFs) based on a wide range of descriptors have been developed. Among those descriptors, protein–ligand interaction fingerprints (IFPs) are competitive due to their simple representations, elaborate profiles of key interactions and easy collaborations with machine-learning algorithms. In this paper, we have adopted a building-block-based taxonomy to review a broad range of IFP models, and compared representative IFP-based SFs in target-specific and generic scoring tasks. Atom-pair-counts-based and substructure-based IFPs show great potential in these tasks.

Unexpected similarity between HIV-1 reverse transcriptase and tumor necrosis factor binding sites revealed by computer vision

Rationalizing the identification of hidden similarities across the repertoire of druggable protein cavities remains a major hurdle to a true proteome-wide structure-based discovery of novel drug candidates. We recently described a new computational approach (ProCare), inspired by numerical image processing, to identify local similarities in fragment-based subpockets. During the validation of the method, we unexpectedly identified a possible similarity in the binding pockets of two unrelated targets, human tumor necrosis factor alpha (TNF-α) and HIV-1 reverse transcriptase (HIV-1 RT). Microscale thermophoresis experiments confirmed the ProCare prediction as two of the three tested and FDA-approved HIV-1 RT inhibitors indeed bind to soluble human TNF-α trimer. Interestingly, the herein disclosed similarity could be revealed neither by state-of-the-art binding sites comparison methods nor by ligand-based pairwise similarity searches, suggesting that the point cloud registration approach implemented in ProCare, is uniquely suited to identify local and unobvious similarities among totally unrelated targets.

DeepPocket: Ligand Binding Site Detection and Segmentation using 3D Convolutional Neural Networks

A structure-based drug design pipeline involves the development of potential drug molecules or ligands that form stable complexes with a given receptor at its binding site. A prerequisite to this is finding druggable and functionally relevant binding sites on the 3D structure of the protein. Although several methods for detecting binding sites have been developed beforehand, a majority of them surprisingly fail in the identification and ranking of binding sites accurately. The rapid adoption and success of deep learning algorithms in various sections of structural biology beckons the usage of such algorithms for accurate binding site detection. As a combination of geometry based software and deep learning, we report a novel framework, DeepPocket that utilizes 3D convolutional neural networks for the rescoring of pockets identified by Fpocket and further segments these identified cavities on the protein surface. Apart from this, we also propose another data set SC6K containing protein structures submitted in the Protein Data Bank (PDB) from January 1st, 2018, until February 28th, 2020, for ligand binding site (LBS) detection. DeepPocket's results on various binding site data sets and SC6K highlight its better performance over current state-of-the-art methods and good generalization ability over novel structures.

Subtype-selective mechanisms of negative allosteric modulators binding to group I metabotropic glutamate receptors

Group I metabotropic glutamate receptors (mGlu₁ and mGlu₅) are promising targets for multiple psychiatric and neurodegenerative disorders. Understanding the subtype selectivity of mGlu₁ and mGlu₅ allosteric sites is essential for the rational design of novel modulators with single- or dual-target mechanism of action. In this study, starting from the deposited mGlu₁ and mGlu₅ crystal structures, we utilized computational modeling approaches integrating docking, molecular dynamics simulation, and efficient post-trajectory analysis to reveal the subtype-selective mechanism of mGlu₁ and mGlu₅ to 10 diverse drug scaffolds representing known negative allosteric modulators (NAMs) in the literature. The results of modeling identified six pairs of non-conserved residues and four pairs of conserved ones as critical features to distinguish the selective NAMs binding to the corresponding receptors. In addition, nine pairs of residues are beneficial to the development of novel dual-target NAMs of group I metabotropic glutamate receptors. Furthermore, the binding modes of a reported dual-target NAM (VU0467558) in mGlu₁ and mGlu₅ were predicted to verify the identified residues that play key roles in the receptor selectivity and the dual-target binding. The results of this study can guide rational structure-based design of novel NAMs, and the approach can be generally applicable to characterize the features of selectivity for other G-protein-coupled receptors.

Exploring the stability of inhibitor binding to SIK2 using molecular dynamics simulation and binding free energy calculation

Salt inducible kinase 2 (SIK2) is a calcium/calmodulin-dependent protein kinase-like kinase that is implicated in a variety of biological phenomena, including cellular metabolism, growth, and apoptosis. SIK2 is the key target for various cancers, including ovarian, breast, prostate, and lung cancers. Although potent inhibitors of SIK2 are being developed, their binding stability and functional role are not presently known. In this work, we studied the detailed interactions between SIK2 and four of its inhibitors, HG-9-91-01, KIN112, MRT67307, and MRT199665, using molecular docking, molecular dynamics simulation, binding free energy calculation, and interaction fingerprint analysis. Intermolecular interactions revealed that HG-9-91-01 and KIN112 have stronger interactions with SIK2 than those of MRT199665 and MRT67307. The key residues involved in binding with SIK2 are conserved among all four inhibitors. Our results explain the detailed interaction of SIK2 with its inhibitors at the molecular level, thus paving the way for the development of targeted efficient anti-cancer drugs.

True Accuracy of Fast Scoring Functions to Predict High-Throughput Screening Data from Docking Poses: The Simpler the Better

Hundreds of fast scoring functions have been developed over the last 20 years to predict binding free energies from three-dimensional structures of protein-ligand complexes. Despite numerous statistical promises, we believe that none of them has been properly validated for daily prospective high-throughput virtual screening studies, mostly because in silico screening challenges usually employ artificially built and biased datasets. We here carry out a fully unbiased evaluation of four scoring functions (Pafnucy, Δ_vinaRF₂₀, IFP, and GRIM) on an in-house developed data collection of experimental high-confidence screening data (LIT-PCBA) covering about 3 million data points on 15 diverse pharmaceutical targets. All four scoring functions were applied to rescore the docking poses of LIT-PCBA compounds in conditions mimicking exactly standard drug discovery scenarios and were compared in terms of propensity to enrich true binders in the top 1%-ranked hit lists. Interestingly, rescoring based on simple interaction fingerprints or interaction graphs outperforms state-of-the-art machine learning and deep learning scoring functions in most of the cases. The current study notably highlights the strong tendency of deep learning methods to predict affinity values within a very narrow range centered on the mean value of samples used for training. Moreover, it suggests that knowledge of pre-existing binding modes is the key to detecting the most potent binders.

Computational representations of protein-ligand interfaces for structure-based virtual screening

Introduction: Structure-based virtual screening (SBVS) is an essential strategy for hit identification. SBVS primarily uses molecular docking, which exploits the protein-ligand binding mode and associated affinity score for compound ranking. Previous studies have shown that computational representation of protein-ligand interfaces and the later establishment of machine learning models are efficacious in improving the accuracy of SBVS.Areas covered: The authors review the computational methods for representing protein-ligand interfaces, which include the traditional ones that use deliberately designed fingerprints and descriptors and the more recent methods that automatically extract features with deep learning. The effects of these methods on the performance of machine learning models are briefly discussed. Additionally, case studies that applied various computational representations to machine learning are cited with remarks.Expert opinion: It has become a trend to extract binding features automatically by deep learning, which uses a completely end-to-end representation. However, there is still plenty of scope for improvement . The interpretability of deep-learning models, the organization of data management, the quantity and quality of available data, and the optimization of hyperparameters could impact the accuracy of feature extraction. In addition, other important structural factors such as water molecules and protein flexibility should be considered.

Understanding the Polypharmacological Profiles of Triple Reuptake Inhibitors by Molecular Simulation

The triple reuptake inhibitors (TRIs) class is a class of effective inhibitors of human monoamine transporters (hMATs), which includes dopamine, norepinephrine, and serotonin transporters (hDATs, hNETs, and hSERTs). Due to the high degree of structural homology of the binding sites of those transporters, it is a great challenge to design potent TRIs with fine-tuned binding profiles. The molecular determinants responsible for the binding selectivity of TRIs to hDATs, hNETs, and hSERTs remain elusive. In this study, the solved X-ray crystallographic structure of hSERT in complex with escitalopram was used as a basis for modeling nine complexes of three representative TRIs (SEP225289, NS2359, and EB1020) bound to their corresponding targets. Molecular dynamics (MD) and effective post-trajectory analysis were performed to estimate the drug binding free energies and characterize the selective profiles of each TRI to hMATs. The common binding mode of studied TRIs to hMATs was revealed by hierarchical clustering analysis of the per-residue energy. Furthermore, the combined protein-ligand interaction fingerprint and residue energy contribution analysis indicated that several conserved and nonconserved "Warm Spots" such as S149, V328, and M427 in hDAT, F317, F323, and V325 in hNET and F335, F341, and V343 in hSERT were responsible for the TRI-binding selectivity. These findings provided important information for rational design of a single drug with better polypharmacological profiles through modulating multiple targets.

Deep Learning in Virtual Screening: Recent Applications and Developments

Drug discovery is a cost and time-intensive process that is often assisted by computational methods, such as virtual screening, to speed up and guide the design of new compounds. For many years, machine learning methods have been successfully applied in the context of computer-aided drug discovery. Recently, thanks to the rise of novel technologies as well as the increasing amount of available chemical and bioactivity data, deep learning has gained a tremendous impact in rational active compound discovery. Herein, recent applications and developments of machine learning, with a focus on deep learning, in virtual screening for active compound design are reviewed. This includes introducing different compound and protein encodings, deep learning techniques as well as frequently used bioactivity and benchmark data sets for model training and testing. Finally, the present state-of-the-art, including the current challenges and emerging problems, are examined and discussed.

Lean-Docking: Exploiting Ligands' Predicted Docking Scores to Accelerate Molecular Docking

In structure-based virtual screening (SBVS), a binding site on a protein structure is used to search for ligands with favorable nonbonded interactions. Because it is computationally difficult, docking is time-consuming and any docking user will eventually encounter a chemical library that is too big to dock. This problem might arise because there is not enough computing power or because preparing and storing so many three-dimensional (3D) ligands requires too much space. In this study, however, we show that quality regressors can be trained to predict docking scores from molecular fingerprints. Although typical docking has a screening rate of less than one ligand per second on one CPU core, our regressors can predict about 5800 docking scores per second. This approach allows us to focus docking on the portion of a database that is predicted to have docking scores below a user-chosen threshold. Herein, usage examples are shown, where only 25% of a ligand database is docked, without any significant virtual screening performance loss. We call this method "lean-docking". To validate lean-docking, a massive docking campaign using several state-of-the-art docking software packages was undertaken on an unbiased data set, with only wet-lab tested active and inactive molecules. Although regressors allow the screening of a larger chemical space, even at a constant docking power, it is also clear that significant progress in the virtual screening power of docking scores is desirable.

SMPLIP-Score: predicting ligand binding affinity from simple and interpretable on-the-fly interaction fingerprint pattern descriptors

In drug discovery, rapid and accurate prediction of protein–ligand binding affinities is a pivotal task for lead optimization with acceptable on-target potency as well as pharmacological efficacy. Furthermore, researchers hope for a high correlation between docking score and pose with key interactive residues, although scoring functions as free energy surrogates of protein–ligand complexes have failed to provide collinearity. Recently, various machine learning or deep learning methods have been proposed to overcome the drawbacks of scoring functions. Despite being highly accurate, their featurization process is complex and the meaning of the embedded features cannot directly be interpreted by human recognition without an additional feature analysis. Here, we propose SMPLIP-Score (Substructural Molecular and Protein–Ligand Interaction Pattern Score), a direct interpretable predictor of absolute binding affinity. Our simple featurization embeds the interaction fingerprint pattern on the ligand-binding site environment and molecular fragments of ligands into an input vectorized matrix for learning layers (random forest or deep neural network). Despite their less complex features than other state-of-the-art models, SMPLIP-Score achieved comparable performance, a Pearson’s correlation coefficient up to 0.80, and a root mean square error up to 1.18 in pK units with several benchmark datasets (PDBbind v.2015, Astex Diverse Set, CSAR NRC HiQ, FEP, PDBbind NMR, and CASF-2016). For this model, generality, predictive power, ranking power, and robustness were examined using direct interpretation of feature matrices for specific targets.

The latest automated docking technologies for novel drug discovery

INTRODUCTION

Molecular docking has been consolidated as one of the most important methods in the molecular modeling field. It has been recognized as a prominent tool in the study of protein-ligand complexes, to describe intermolecular interactions, to accurately predict poses of multiple ligands, to discover novel promising bioactive compounds. Molecular docking methods have evolved in terms of their accuracy and reliability; but there are pending issues to solve for improving the connection between the docking results and the experimental evidence.

AREAS COVERED

In this article, the author reviews very recent innovative molecular docking applications with special emphasis on reverse docking, treatment of protein flexibility, the use of experimental data to guide the selection of docking poses, the application of Quantum mechanics(QM) in docking, and covalent docking.

EXPERT OPINION

There are several issues being worked on in recent years that will lead to important breakthroughs in molecular docking methods in the near future These developments are related to more efficient exploration of large datasets and receptor conformations, advances in electronic description, and the use of structural information for guiding the selection of results.

Structural Insights into the Binding Modes of Viral RNA-Dependent RNA Polymerases Using a Function-Site Interaction Fingerprint Method for RNA Virus Drug Discovery

The coronavirus disease of 2019 (COVID-19) pandemic speaks to the need for drugs that not only are effective but also remain effective given the mutation rate of severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2). To this end, we describe structural binding-site insights for facilitating COVID-19 drug design when targeting RNA-dependent RNA polymerase (RDRP), a common conserved component of RNA viruses. We combined an RDRP structure data set, including 384 RDRP PDB structures and all corresponding RDRP-ligand interaction fingerprints, thereby revealing the structural characteristics of the active sites for application to RDRP-targeted drug discovery. Specifically, we revealed the intrinsic ligand-binding modes and associated RDRP structural characteristics. Four types of binding modes with corresponding binding pockets were determined, suggesting two major subpockets available for drug discovery. We screened a drug data set of 7894 compounds against these binding pockets and presented the top-10 small molecules as a starting point in further exploring the prevention of virus replication. In summary, the binding characteristics determined here help rationalize RDRP-targeted drug discovery and provide insights into the specific binding mechanisms important for containing the SARS-CoV-2 virus.

Merging Ligand-Based and Structure-Based Methods in Drug Discovery: An Overview of Combined Virtual Screening Approaches

Virtual screening (VS) is an outstanding cornerstone in the drug discovery pipeline. A variety of computational approaches, which are generally classified as ligand-based (LB) and structure-based (SB) techniques, exploit key structural and physicochemical properties of ligands and targets to enable the screening of virtual libraries in the search of active compounds. Though LB and SB methods have found widespread application in the discovery of novel drug-like candidates, their complementary natures have stimulated continued efforts toward the development of hybrid strategies that combine LB and SB techniques, integrating them in a holistic computational framework that exploits the available information of both ligand and target to enhance the success of drug discovery projects. In this review, we analyze the main strategies and concepts that have emerged in the last years for defining hybrid LB + SB computational schemes in VS studies. Particularly, attention is focused on the combination of molecular similarity and docking, illustrating them with selected applications taken from the literature.

Selective Inhibition of HDAC1 by Macrocyclic Polypeptide for the Treatment of Glioblastoma: A Binding Mechanistic Analysis Based on Molecular Dynamics

Glioblastoma (GBM) is the most common and aggressive intracranial malignant brain tumor, and the abnormal expression of HDAC1 is closely correlated to the progression, recurrence and metastasis of GBM cells, making selective inhibition of HDAC1 a promising strategy for GBM treatments. Among all available selective HDAC1 inhibitors, the macrocyclic peptides have gained great attention due to their remarkable inhibitory selectivity on HDAC1. However, the binding mechanism underlying this selectivity is still elusive, which increases the difficulty of designing and synthesizing the macrocyclic peptide-based anti-GBM drug. Herein, multiple computational approaches were employed to explore the binding behaviors of a typical macrocyclic peptide FK228 in both HDAC1 and HDAC6. Starting from the docking conformations of FK228 in the binding pockets of HDAC1&6, relatively long MD simulation (500 ns) shown that the hydrophobic interaction and hydrogen bonding of E91 and D92 in the Loop2 of HDAC1 with the Cap had a certain traction effect on FK228, and the sub-pocket formed by Loop1 and Loop2 in HDAC1 could better accommodate the Cap group, which had a positive effect on maintaining the active conformation of FK228. While the weakening of the interactions between FK228 and the residues in the Loop2 of HDAC6 during the MD simulation led to the large deflection of FK228 in the binding site, which also resulted in the decrease in the interactions between the Linker region of FK228 and the previously identified key amino acids (H134, F143, H174, and F203). Therefore, the residues located in Loop1 and Loop2 contributed in maintaining the active conformation of FK228, which would provide valuable hints for the discovery and design of novel macrocyclic polypeptide HDAC inhibitors.