Human-in-the-loop assisted de novo molecular design.

A de novo molecular design workflow can be used together with technologies such as reinforcement learning to navigate the chemical space. A bottleneck in the workflow that remains to be solved is how to integrate human feedback in the exploration of the chemical space to optimize molecules. A human drug designer still needs to design the goal, expressed as a scoring function for the molecules that captures the designer's implicit knowledge about the optimization task. Little support for this task exists and, consequently, a chemist usually resorts to iteratively building the objective function of multi-parameter optimization (MPO) in de novo design. We propose a principled approach to use human-in-the-loop machine learning to help the chemist to adapt the MPO scoring function to better match their goal. An advantage is that the method can learn the scoring function directly from the user's feedback while they browse the output of the molecule generator, instead of the current manual tuning of the scoring function with trial and error. The proposed method uses a probabilistic model that captures the user's idea and uncertainty about the scoring function, and it uses active learning to interact with the user. We present two case studies for this: In the first use-case, the parameters of an MPO are learned, and in the second use-case a non-parametric component of the scoring function to capture human domain knowledge is developed. The results show the effectiveness of the methods in two simulated example cases with an oracle, achieving significant improvement in less than 200 feedback queries, for the goals of a high QED score and identifying potent molecules for the DRD2 receptor, respectively. We further demonstrate the performance gains with a medicinal chemist interacting with the system.

Icolos: a workflow manager for structure-based post-processing of de novo generated small molecules.

SUMMARY: We present Icolos, a workflow manager written in Python as a tool for automating complex structure-based workflows for drug design. Icolos can be used as a standalone tool, for example in virtual screening campaigns, or can be used in conjunction with deep learning-based molecular generation facilitated for example by REINVENT, a previously published molecular de novo design package. In this publication, we focus on the internal structure and general capabilities of Icolos, using molecular docking experiments as an illustrative example. AVAILABILITY AND IMPLEMENTATION: The source code is freely available at https://github.com/MolecularAI/Icolos under the Apache 2.0 license. Tutorial notebooks containing minimal working examples can be found at https://github.com/MolecularAI/IcolosCommunity. SUPPLEMENTARY INFORMATION: Supplementary data are available at Bioinformatics online.

Transformer-based molecular optimization beyond matched molecular pairs.

Molecular optimization aims to improve the drug profile of a starting molecule. It is a fundamental problem in drug discovery but challenging due to (i) the requirement of simultaneous optimization of multiple properties and (ii) the large chemical space to explore. Recently, deep learning methods have been proposed to solve this task by mimicking the chemist's intuition in terms of matched molecular pairs (MMPs). Although MMPs is a widely used strategy by medicinal chemists, it offers limited capability in terms of exploring the space of structural modifications, therefore does not cover the complete space of solutions. Often more general transformations beyond the nature of MMPs are feasible and/or necessary, e.g. simultaneous modifications of the starting molecule at different places including the core scaffold. This study aims to provide a general methodology that offers more general structural modifications beyond MMPs. In particular, the same Transformer architecture is trained on different datasets. These datasets consist of a set of molecular pairs which reflect different types of transformations. Beyond MMP transformation, datasets reflecting general structural changes are constructed from ChEMBL based on two approaches: Tanimoto similarity (allows for multiple modifications) and scaffold matching (allows for multiple modifications but keep the scaffold constant) respectively. We investigate how the model behavior can be altered by tailoring the dataset while using the same model architecture. Our results show that the models trained on differently prepared datasets transform a given starting molecule in a way that it reflects the nature of the dataset used for training the model. These models could complement each other and unlock the capability for the chemists to pursue different options for improving a starting molecule.

Structure-Aware Multimodal Deep Learning for Drug-Protein Interaction Prediction.

Identifying drug-protein interactions (DPIs) is crucial in drug discovery, and a number of machine learning methods have been developed to predict DPIs. Existing methods usually use unrealistic data sets with hidden bias, which will limit the accuracy of virtual screening methods. Meanwhile, most DPI prediction methods pay more attention to molecular representation but lack effective research on protein representation and high-level associations between different instances. To this end, we present the novel structure-aware multimodal deep DPI prediction model, STAMP-DPI, which was trained on a curated industry-scale benchmark data set. We built a high-quality benchmark data set named GalaxyDB for DPI prediction. This industry-scale data set along with an unbiased training procedure resulted in a more robust benchmark study. For informative protein representation, we constructed a structure-aware graph neural network method from the protein sequence by combining predicted contact maps and graph neural networks. Through further integration of structure-based representation and high-level pretrained embeddings for molecules and proteins, our model effectively captures the feature representation of the interactions between them. As a result, STAMP-DPI outperformed state-of-the-art DPI prediction methods by decreasing 7.00% mean square error (MSE) in the Davis data set and improving 8.89% area under the curve (AUC) in the GalaxyDB data set. Moreover, our model is an interpretable model with the transformer-based interaction mechanism, which can accurately reveal the binding sites between molecules and proteins.

LibINVENT: Reaction-based Generative Scaffold Decoration for in Silico Library Design.

Because of the strong relationship between the desired molecular activity and its structural core, the screening of focused, core-sharing chemical libraries is a key step in lead optimization. Despite the plethora of current research focused on in silico methods for molecule generation, to our knowledge, no tool capable of designing such libraries has been proposed. In this work, we present a novel tool for de novo drug design called LibINVENT. It is capable of rapidly proposing chemical libraries of compounds sharing the same core while maximizing a range of desirable properties. To further help the process of designing focused libraries, the user can list specific chemical reactions that can be used for the library creation. LibINVENT is therefore a flexible tool for generating virtual chemical libraries for lead optimization in a broad range of scenarios. Additionally, the shared core ensures that the compounds in the library are similar, possess desirable properties, and can also be synthesized under the same or similar conditions. The LibINVENT code is freely available in our public repository at https://github.com/MolecularAI/Lib-INVENT. The code necessary for data preprocessing is further available at: https://github.com/MolecularAI/Lib-INVENT-dataset.

Has Artificial Intelligence Impacted Drug Discovery?

Artificial intelligence (AI) tools find increasing application in drug discovery supporting every stage of the Design-Make-Test-Analyse (DMTA) cycle. The main focus of this chapter is the application in molecular generation with the aid of deep neural networks (DNN). We present a historical overview of the main advances in the field. We analyze the concepts of distribution and goal-directed learning and then highlight some of the recent applications of generative models in drug design with a focus into research work from the biopharmaceutical industry. We present in some more detail REINVENT which is an open-source software developed within our group in AstraZeneca and the main platform for AI molecular design support for a number of medicinal chemistry projects in the company and we also demonstrate some of our work in library design. Finally, we present some of the main challenges in the application of AI in Drug Discovery and different approaches to respond to these challenges which define areas for current and future work.

DockStream: a docking wrapper to enhance de novo molecular design.

Recently, we have released the de novo design platform REINVENT in version 2.0. This improved and extended iteration supports far more features and scoring function components, which allows bespoke and tailor-made protocols to maximize impact in small molecule drug discovery projects. A major obstacle of generative models is producing active compounds, in which predictive (QSAR) models have been applied to enrich target activity. However, QSAR models are inherently limited by their applicability domains. To overcome these limitations, we introduce a structure-based scoring component for REINVENT. DockStream is a flexible, stand-alone molecular docking wrapper that provides access to a collection of ligand embedders and docking backends. Using the benchmarking and analysis workflow provided in DockStream, execution and subsequent analysis of a variety of docking configurations can be automated. Docking algorithms vary greatly in performance depending on the target and the benchmarking and analysis workflow provides a streamlined solution to identifying productive docking configurations. We show that an informative docking configuration can inform the REINVENT agent to optimize towards improving docking scores using public data. With docking activated, REINVENT is able to retain key interactions in the binding site, discard molecules which do not fit the binding cavity, harness unused (sub-)pockets, and improve overall performance in the scaffold-hopping scenario. The code is freely available at https://github.com/MolecularAI/DockStream .

De novo design with deep generative models based on 3D similarity scoring.

We have demonstrated the utility of a 3D shape and pharmacophore similarity scoring component in molecular design with a deep generative model trained with reinforcement learning. Using Dopamine receptor type 2 (DRD2) as an example and its antagonist haloperidol 1 as a starting point in a ligand based design context, we have shown in a retrospective study that a 3D similarity enabled generative model can discover new leads in the absence of any other information. It can be efficiently used for scaffold hopping and generation of novel series. 3D similarity based models were compared against 2D QSAR based, indicating a significant degree of orthogonality of the generated outputs and with the former having a more diverse output. In addition, when the two scoring components are combined together for training of the generative model, it results in more efficient exploration of desirable chemical space compared to the individual components.

REINVENT 2.0: An AI Tool for De Novo Drug Design.

In the past few years, we have witnessed a renaissance of the field of molecular de novo drug design. The advancements in deep learning and artificial intelligence (AI) have triggered an avalanche of ideas on how to translate such techniques to a variety of domains including the field of drug design. A range of architectures have been devised to find the optimal way of generating chemical compounds by using either graph- or string (SMILES)-based representations. With this application note, we aim to offer the community a production-ready tool for de novo design, called REINVENT. It can be effectively applied on drug discovery projects that are striving to resolve either exploration or exploitation problems while navigating the chemical space. It can facilitate the idea generation process by bringing to the researcher's attention the most promising compounds. REINVENT's code is publicly available at https://github.com/MolecularAI/Reinvent.

SMILES-based deep generative scaffold decorator for de-novo drug design.

Molecular generative models trained with small sets of molecules represented as SMILES strings can generate large regions of the chemical space. Unfortunately, due to the sequential nature of SMILES strings, these models are not able to generate molecules given a scaffold (i.e., partially-built molecules with explicit attachment points). Herein we report a new SMILES-based molecular generative architecture that generates molecules from scaffolds and can be trained from any arbitrary molecular set. This approach is possible thanks to a new molecular set pre-processing algorithm that exhaustively slices all possible combinations of acyclic bonds of every molecule, combinatorically obtaining a large number of scaffolds with their respective decorations. Moreover, it serves as a data augmentation technique and can be readily coupled with randomized SMILES to obtain even better results with small sets. Two examples showcasing the potential of the architecture in medicinal and synthetic chemistry are described: First, models were trained with a training set obtained from a small set of Dopamine Receptor D2 (DRD2) active modulators and were able to meaningfully decorate a wide range of scaffolds and obtain molecular series predicted active on DRD2. Second, a larger set of drug-like molecules from ChEMBL was selectively sliced using synthetic chemistry constraints (RECAP rules). In this case, the resulting scaffolds with decorations were filtered only to allow those that included fragment-like decorations. This filtering process allowed models trained with this dataset to selectively decorate diverse scaffolds with fragments that were generally predicted to be synthesizable and attachable to the scaffold using known synthetic approaches. In both cases, the models were already able to decorate molecules using specific knowledge without the need to add it with other techniques, such as reinforcement learning. We envision that this architecture will become a useful addition to the already existent architectures for de novo molecular generation.

A combined in silico and in vitro study on mouse Serpina1a antitrypsin-deficiency mutants.

Certain point-mutations in the human SERPINA1-gene can cause severe α1-antitrypsin-deficiency (A1AT-D). Affected individuals can suffer from loss-of-function lung-disease and from gain-of-function liver-disease phenotypes. However, age of onset and severity of clinical appearance is heterogeneous amongst carriers, suggesting involvement of additional genetic and environmental factors. The generation of authentic A1AT-D mouse-models has been hampered by the complexity of the mouse Serpina1-gene locus and a model with concurrent lung and liver-disease is still missing. Here, we investigate point-mutations in the mouse Serpina1a antitrypsin-orthologue, which are homolog-equivalent to ones known to cause severe A1AT-D in human. We combine in silico and in vitro methods and we find that analyzed mutations do introduce potential disease-causing properties into Serpina1a. Finally, we show that introduction of the King's-mutation causes inactivation of neutrophil elastase inhibitory-function in both, mouse and human antitrypsin, while the mouse Z-mutant retains activity. This work paves the path to generation of better A1AT-D mouse-models.

BiPPred: Combined sequence- and structure-based prediction of peptide binding to the Hsp70 chaperone BiP.

Substrate binding to Hsp70 chaperones is involved in many biological processes, and the identification of potential substrates is important for a comprehensive understanding of these events. We present a multi-scale pipeline for an accurate, yet efficient prediction of peptides binding to the Hsp70 chaperone BiP by combining sequence-based prediction with molecular docking and MMPBSA calculations. First, we measured the binding of 15mer peptides from known substrate proteins of BiP by peptide array (PA) experiments and performed an accuracy assessment of the PA data by fluorescence anisotropy studies. Several sequence-based prediction models were fitted using this and other peptide binding data. A structure-based position-specific scoring matrix (SB-PSSM) derived solely from structural modeling data forms the core of all models. The matrix elements are based on a combination of binding energy estimations, molecular dynamics simulations, and analysis of the BiP binding site, which led to new insights into the peptide binding specificities of the chaperone. Using this SB-PSSM, peptide binders could be predicted with high selectivity even without training of the model on experimental data. Additional training further increased the prediction accuracies. Subsequent molecular docking (DynaDock) and MMGBSA/MMPBSA-based binding affinity estimations for predicted binders allowed the identification of the correct binding mode of the peptides as well as the calculation of nearly quantitative binding affinities. The general concept behind the developed multi-scale pipeline can readily be applied to other protein-peptide complexes with linearly bound peptides, for which sufficient experimental binding data for the training of classical sequence-based prediction models is not available. Proteins 2016; 84:1390-1407. © 2016 Wiley Periodicals, Inc.

Multiple sclerosis: Molecular mimicry of an antimyelin HLA class I restricted T-cell receptor.

OBJECTIVE: To identify target antigens presented by human leukocyte antigen (HLA)-A*02:01 to the myelin-reactive human T-cell receptor (TCR) 2D1, which was originally isolated from a CD8+ T-cell clone recognizing proteolipid protein (PLP) in the context of HLA-A*03:01, we employed a new antigen search technology. METHODS: We used our recently developed antigen search technology that employs plasmid-encoded combinatorial peptide libraries and a highly sensitive single cell detection system to identify endogenous candidate peptides of mice and human origin. We validated candidate antigens by independent T-cell assays using synthetic peptides and refolded HLA:peptide complexes. A molecular model of HLA-A*02:01:peptide complexes was obtained by molecular dynamics simulations. RESULTS: We identified one peptide from glycerolphosphatidylcholine phosphodiesterase 1, which is identical in mice and humans and originates from a protein that is expressed in many cell types. When bound to HLA-A*02:01, this peptide cross-stimulates the PLP-reactive HLA-A3-restricted TCR 2D1. Investigation of molecular details revealed that the peptide length plays a crucial role in its capacity to bind HLA-A*02:01 and to activate TCR 2D1. Molecular modeling illustrated the 3D structures of activating HLA:peptide complexes. CONCLUSIONS: Our results show that our antigen search technology allows us to identify new candidate antigens of a presumably pathogenic, autoreactive, human CD8+ T-cell-derived TCR. They further illustrate how this TCR, which recognizes a myelin peptide bound to HLA-A*03:01, may cross-react with an unrelated peptide presented by the protective HLA class I allele HLA-A*02:01.

A Cohesive and Integrated Platform for Immunogenicity Prediction.

In silico methods for immunogenicity prediction mine the enormous quantity of data arising from deciphered genomes and proteomes to identify immunogenic proteins. While high and productive immunogenicity is essential for vaccines, therapeutic proteins and monoclonal antibodies should be minimally immunogenic. Here, we present a cohesive platform for immunogenicity and MHC class I and/or II binding affinity prediction. The platform integrates three quasi-independent modular servers: VaxiJen, EpiJen, and EpiTOP. VaxiJen (http://www.ddg-pharmfac.net/vaxijen) predicts immunogenicity of proteins of different origin; EpiJen (http://www.ddg-pharmfac.net/epijen) predicts peptide binding to MHC class I proteins; and EpiTOP (http://www.ddg-pharmfac.net/epitop) predicts peptide binding to MHC class II proteins. The platform is freely accessible and user-friendly. The protocol for immunogenicity prediction is demonstrated by selecting immunogenic proteins from Mycobacterium tuberculosis and predicting how the peptide epitopes within them bind to MHC class I and class II proteins.

4-Fluoro-3',4',5'-trimethoxychalcone as a new anti-invasive agent. From discovery to initial validation in an in vivo metastasis model.

Invasion and metastasis are responsible for 90% of cancer-related mortality. Herein, we report on our quest for novel, clinically relevant inhibitors of local invasion, based on a broad screen of natural products in a phenotypic assay. Starting from micromolar chalcone hits, a predictive QSAR model for diaryl propenones was developed, and synthetic analogues with a 100-fold increase in potency were obtained. Two nanomolar hits underwent efficacy validation and eADMET profiling; one compound was shown to increase the survival time in an artificial metastasis model in nude mice. Although the molecular mechanism(s) by which these substances mediate efficacy remain(s) unrevealed, we were able to eliminate the major targets commonly associated with antineoplastic chalcones.

Histidine hydrogen bonding in MHC at pH 5 and pH 7 modeled by molecular docking and molecular dynamics simulations.

Hydrogen bonds play important roles in maintaining the structure of proteins and in the formation of most biomolecular protein-ligand complexes. All amino acids can act as hydrogen bond donors and acceptors. Among amino acids, Histidine is unique, as it can exist in neutral or positively charged forms within the physiological pH range of 5.0 to 7.0. Histidine can thus interact with other aromatic residues as well as forming hydrogen bonds with polar and charged residues. The ability of His to exchange a proton lies at the heart of many important functional biomolecular interactions, including immunological ones. By using molecular docking and molecular dynamics simulation, we examine the influence of His protonation/deprotonation on peptide binding affinity to MHC class II proteins from locus HLA-DP. Peptide-MHC interaction underlies the adaptive cellular immune response, upon which the next generation of commercially-important vaccines will depend. Consistent with experiment, we find that peptides containing protonated His residues bind better to HLA-DP proteins than those with unprotonated His. Enhanced binding at pH 5.0 is due, in part, to additional hydrogen bonds formed between peptide His(+) and DP proteins. In acidic endosomes, protein His(79ß) is predominantly protonated. As a result, the peptide binding cleft narrows in the vicinity of His(79ß), which stabilizes the peptide - HLA-DP protein complex.

Immunoinformatic docking approach for the analysis of KIR3DL1/HLA-B interaction.

KIR3DL1 is among the most interesting receptors studied, within the killer immunoglobulin receptor (KIR) family. Human leukocyte antigen (HLA) class I Bw4 epitope inhibits strongly Natural Killer (NK) cell's activity through interaction with KIR3DL1 receptor, while Bw6 generally does not. This interaction has been indicated to play an important role in the immune control of different viral infectious diseases. However, the structural interaction between the KIR3DL1 receptor and different HLA-B alleles has been scarcely studied. To understand the complexity of KIR3DL1-HLA-B interaction, HLA-B alleles carrying Bw4/Bw6 epitope and KIR3DL1∗001 allele in presence of different peptides has been evaluated by using a structural immunoinformatic approach. Different energy minimization force fields (ff) have been tested and NOVA ff enables the successful prediction of ligand-receptor interaction. HLA-B alleles carrying Bw4 epitope present the highest capability of interaction with KIR3DL1∗001 compared to the HLA-B alleles presenting Bw6. The presence of the epitope Bw4 determines a conformational change which leads to a stronger interaction between nonpolymorphic arginine at position 79 of HLA-B and KIR3DL1∗001 136-142 loop. The data shed new light on the modalities of KIR3DL1 interaction with HLA-B alleles essential for the modulation of NK immune-mediated response.

EpiDOCK: a molecular docking-based tool for MHC class II binding prediction.

Cellular peptide vaccines contain T-cell epitopes. The main prerequisite for a peptide to act as a T-cell epitope is that it binds to a major histocompatibility complex (MHC) protein. Peptide MHC binder identification is an extremely costly experimental challenge since human MHCs, named human leukocyte antigen, are highly polymorphic and polygenic. Here we present EpiDOCK, the first structure-based server for MHC class II binding prediction. EpiDOCK predicts binding to the 23 most frequent human, MHC class II proteins. It identifies 90% of true binders and 76% of true non-binders, with an overall accuracy of 83%. EpiDOCK is freely accessible at http://epidock.ddg-pharmfac.net.

T-cell epitope vaccine design by immunoinformatics.

Vaccination is generally considered to be the most effective method of preventing infectious diseases. All vaccinations work by presenting a foreign antigen to the immune system in order to evoke an immune response. The active agent of a vaccine may be intact but inactivated ('attenuated') forms of the causative pathogens (bacteria or viruses), or purified components of the pathogen that have been found to be highly immunogenic. The increased understanding of antigen recognition at molecular level has resulted in the development of rationally designed peptide vaccines. The concept of peptide vaccines is based on identification and chemical synthesis of B-cell and T-cell epitopes which are immunodominant and can induce specific immune responses. The accelerating growth of bioinformatics techniques and applications along with the substantial amount of experimental data has given rise to a new field, called immunoinformatics. Immunoinformatics is a branch of bioinformatics dealing with in silico analysis and modelling of immunological data and problems. Different sequence- and structure-based immunoinformatics methods are reviewed in the paper.