Kinome-Wide Profiling Prediction of Small Molecules

Databases of ligand-binding pockets and protein-ligand interactions

Many research groups and institutions have created a variety of databases curating experimental and predicted data related to protein-ligand binding. The landscape of available databases is dynamic, with new databases emerging and established databases becoming defunct. Here, we review the current state of databases that contain binding pockets and protein-ligand binding interactions. We have compiled a list of such databases, fifty-three of which are currently available for use. We discuss variation in how binding pockets are defined and summarize pocket-finding methods. We organize the fifty-three databases into subgroups based on goals and contents, and describe standard use cases. We also illustrate that pockets within the same protein are characterized differently across different databases. Finally, we assess critical issues of sustainability, accessibility and redundancy.

Prediction of viral protease inhibitors using proteochemometrics approach

Being widely accepted tools in computational drug search, the (Q)SAR methods have limitations related to data incompleteness. The proteochemometrics (PCM) approach expands the applicability area by using description for both protein and ligand structures. The PCM algorithms are urgently required for the development of new antiviral agents. We suggest the PCM method using the TLMNA descriptors, combining the MNA descriptors of ligands and protein sequence N-grams. Our method was validated on the viral chymotrypsin-like proteases and their ligands. We have developed an original protocol allowing us to collect a comprehensive set of 15 protein sequences and more than 9000 ligands from the ChEMBL database. The N-grams were derived from the 3D-based alignment, accurately superposing ligand-binding regions. In testing the ligand set in SAR mode with MNA descriptors, an accuracy above 0.95 was determined that shows the perspective of the antiviral drug search in virtual chemical libraries. The effective PCM models were built with the TLMNA descriptor. The strong validation procedure with pair exclusion simulated the prediction of interactions between the new ligands and new targets, resulting in accuracy estimation up to 0.89. The PCM approach shows slightly lower accuracy caused by more uncertainty compared with SAR, but it overcomes the problem of data incompleteness.

Large-scale comparison of machine learning methods for profiling prediction of kinase inhibitors

Conventional machine learning (ML) and deep learning (DL) play a key role in the selectivity prediction of kinase inhibitors. A number of models based on available datasets can be used to predict the kinase profile of compounds, but there is still controversy about the advantages and disadvantages of ML and DL for such tasks. In this study, we constructed a comprehensive benchmark dataset of kinase inhibitors, involving in 141,086 unique compounds and 216,823 well-defined bioassay data points for 354 kinases. We then systematically compared the performance of 12 ML and DL methods on the kinase profiling prediction task. Extensive experimental results reveal that (1) Descriptor-based ML models generally slightly outperform fingerprint-based ML models in terms of predictive performance. RF as an ensemble learning approach displays the overall best predictive performance. (2) Single-task graph-based DL models are generally inferior to conventional descriptor- and fingerprint-based ML models, however, the corresponding multi-task models generally improves the average accuracy of kinase profile prediction. For example, the multi-task FP-GNN model outperforms the conventional descriptor- and fingerprint-based ML models with an average AUC of 0.807. (3) Fusion models based on voting and stacking methods can further improve the performance of the kinase profiling prediction task, specifically, RF::AtomPairs + FP2 + RDKitDes fusion model performs best with the highest average AUC value of 0.825 on the test sets. These findings provide useful information for guiding choices of the ML and DL methods for the kinase profiling prediction tasks. Finally, an online platform called KIPP ( https://kipp.idruglab.cn ) and python software are developed based on the best models to support the kinase profiling prediction, as well as various kinase inhibitor identification tasks including virtual screening, compound repositioning and target fishing.

Redesigning plant specialized metabolism with supervised machine learning using publicly available reactome data

The immense structural diversity of products and intermediates of plant specialized metabolism (specialized metabolites) makes them rich sources of therapeutic medicine, nutrients, and other useful materials. With the rapid accumulation of reactome data that can be accessible on biological and chemical databases, along with recent advances in machine learning, this review sets out to outline how supervised machine learning can be used to design new compounds and pathways by exploiting the wealth of said data. We will first examine the various sources from which reactome data can be obtained, followed by explaining the different machine learning encoding methods for reactome data. We then discuss current supervised machine learning developments that can be employed in various aspects to help redesign plant specialized metabolism.

Artificial intelligence to guide precision anticancer therapy with multitargeted kinase inhibitors

BACKGROUND

Vast amounts of rapidly accumulating biological data related to cancer and a remarkable progress in the field of artificial intelligence (AI) have paved the way for precision oncology. Our recent contribution to this area of research is CancerOmicsNet, an AI-based system to predict the therapeutic effects of multitargeted kinase inhibitors across various cancers. This approach was previously demonstrated to outperform other deep learning methods, graph kernel models, molecular docking, and drug binding pocket matching.

METHODS

CancerOmicsNet integrates multiple heterogeneous data by utilizing a deep graph learning model with sophisticated attention propagation mechanisms to extract highly predictive features from cancer-specific networks. The AI-based system was devised to provide more accurate and robust predictions than data-driven therapeutic discovery using gene signature reversion.

RESULTS

Selected CancerOmicsNet predictions obtained for "unseen" data are positively validated against the biomedical literature and by live-cell time course inhibition assays performed against breast, pancreatic, and prostate cancer cell lines. Encouragingly, six molecules exhibited dose-dependent antiproliferative activities, with pan-CDK inhibitor JNJ-7706621 and Src inhibitor PP1 being the most potent against the pancreatic cancer cell line Panc 04.03.

CONCLUSIONS

CancerOmicsNet is a promising AI-based platform to help guide the development of new approaches in precision oncology involving a variety of tumor types and therapeutics.

CancerOmicsNet: a multi-omics network-based approach to anti-cancer drug profiling

Development of novel anti-cancer treatments requires not only a comprehensive knowledge of cancer processes and drug mechanisms of action, but also the ability to accurately predict the response of various cancer cell lines to therapeutics. Numerous computational methods have been developed to address this issue, including algorithms employing supervised machine learning. Nonetheless, high prediction accuracies reported for many of these techniques may result from a significant overlap among training, validation, and testing sets, making existing predictors inapplicable to new data. To address these issues, we developed CancerOmicsNet, a graph neural network with sophisticated attention propagation mechanisms to predict the therapeutic effects of kinase inhibitors across various tumors. Emphasizing on the system-level complexity of cancer, CancerOmicsNet integrates multiple heterogeneous data, such as biological networks, genomics, inhibitor profiling, and gene-disease associations, into a unified graph structure. The performance of CancerOmicsNet, properly cross-validated at the tissue level, is 0.83 in terms of the area under the receiver operating characteristics, which is notably higher than those measured for other approaches. CancerOmicsNet generalizes well to unseen data, i.e., it can predict therapeutic effects across a variety of cancer cell lines and inhibitors. CancerOmicsNet is freely available to the academic community at https://github.com/pulimeng/CancerOmicsNet.

An integrated network representation of multiple cancer-specific data for graph-based machine learning

Genomic profiles of cancer cells provide valuable information on genetic alterations in cancer. Several recent studies employed these data to predict the response of cancer cell lines to drug treatment. Nonetheless, due to the multifactorial phenotypes and intricate mechanisms of cancer, the accurate prediction of the effect of pharmacotherapy on a specific cell line based on the genetic information alone is problematic. Emphasizing on the system-level complexity of cancer, we devised a procedure to integrate multiple heterogeneous data, including biological networks, genomics, inhibitor profiling, and gene-disease associations, into a unified graph structure. In order to construct compact, yet information-rich cancer-specific networks, we developed a novel graph reduction algorithm. Driven by not only the topological information, but also the biological knowledge, the graph reduction increases the feature-only entropy while preserving the valuable graph-feature information. Subsequent comparative benchmarking simulations employing a tissue level cross-validation protocol demonstrate that the accuracy of a graph-based predictor of the drug efficacy is 0.68, which is notably higher than those measured for more traditional, matrix-based techniques on the same data. Overall, the non-Euclidean representation of the cancer-specific data improves the performance of machine learning to predict the response of cancer to pharmacotherapy. The generated data are freely available to the academic community at https://osf.io/dzx7b/.

Sequence-based prediction of protein binding regions and drug-target interactions

Identifying drug-target interactions (DTIs) is important for drug discovery. However, searching all drug-target spaces poses a major bottleneck. Therefore, recently many deep learning models have been proposed to address this problem. However, the developers of these deep learning models have neglected interpretability in model construction, which is closely related to a model's performance. We hypothesized that training a model to predict important regions on a protein sequence would increase DTI prediction performance and provide a more interpretable model. Consequently, we constructed a deep learning model, named Highlights on Target Sequences (HoTS), which predicts binding regions (BRs) between a protein sequence and a drug ligand, as well as DTIs between them. To train the model, we collected complexes of protein-ligand interactions and protein sequences of binding sites and pretrained the model to predict BRs for a given protein sequence-ligand pair via object detection employing transformers. After pretraining the BR prediction, we trained the model to predict DTIs from a compound token designed to assign attention to BRs. We confirmed that training the BRs prediction model indeed improved the DTI prediction performance. The proposed HoTS model showed good performance in BR prediction on independent test datasets even though it does not use 3D structure information in its prediction. Furthermore, the HoTS model achieved the best performance in DTI prediction on test datasets. Additional analysis confirmed the appropriate attention for BRs and the importance of transformers in BR and DTI prediction. The source code is available on GitHub ( https://github.com/GIST-CSBL/HoTS ).

Generalized vec trick for fast learning of pairwise kernel models

Pairwise learning corresponds to the supervised learning setting where the goal is to make predictions for pairs of objects. Prominent applications include predicting drug-target or protein-protein interactions, or customer-product preferences. In this work, we present a comprehensive review of pairwise kernels, that have been proposed for incorporating prior knowledge about the relationship between the objects. Specifically, we consider the standard, symmetric and anti-symmetric Kronecker product kernels, metric-learning, Cartesian, ranking, as well as linear, polynomial and Gaussian kernels. Recently, a $$O(nm+nq)$$ O ( n m + n q ) time generalized vec trick algorithm, where $$n$$ n , $$m$$ m , and $$q$$ q denote the number of pairs, drugs and targets, was introduced for training kernel methods with the Kronecker product kernel. This was a significant improvement over previous $$O(n^2)$$ O ( n 2 ) training methods, since in most real-world applications $$m,q<< n$$ m , q < < n . In this work we show how all the reviewed kernels can be expressed as sums of Kronecker products, allowing the use of generalized vec trick for speeding up their computation. In the experiments, we demonstrate how the introduced approach allows scaling pairwise kernels to much larger data sets than previously feasible, and provide an extensive comparison of the kernels on a number of biological interaction prediction tasks.

S2DV: converting SMILES to a drug vector for predicting the activity of anti-HBV small molecules

In the past few decades, chronic hepatitis B caused by hepatitis B virus (HBV) has been one of the most serious diseases to human health. The development of innovative systems is essential for preventing the complex pathogenesis of hepatitis B and reducing side effects caused by drugs. HBV inhibitory drugs have been developed through various compounds, and they are often limited by routine experimental screening and delay drug development. More recently, virtual screening of compounds has gradually been used in drug research with strong computational capability and is further applied in anti-HBV drug screening, thus facilitating a reliable drug screening process. However, the lack of structural information in traditional compound analysis is an important hurdle for unsatisfactory efficiency in drug screening. Here, a natural language processing technique was adopted to analyze compound simplified molecular input line entry system strings. By using the targeted optimized word2vec model for pretraining, we can accurately represent the relationship between the compound and its substructure. The machine learning model based on training results can effectively predict the inhibitory effect of compounds on HBV and liver toxicity. The reliability of the model is verified by the results of wet-lab experiments. In addition, a tool has been published to predict potential compounds. Hence, this article provides a new perspective on the prediction of compound properties for anti-HBV drugs that can help improve hepatitis B diagnosis and further develop human health in the future.

Exploration and augmentation of pharmacological space via adversarial auto-encoder model for facilitating kinase-centric drug development

Predicting compound-protein interactions (CPIs) is of great importance for drug discovery and repositioning, yet still challenging mainly due to the sparse nature of CPI matrixes, resulting in poor generalization performance. Hence, unlike typical CPI prediction models focused on representation learning or model selection, we propose a deep neural network-based strategy, PCM-AAE, that re-explores and augments the pharmacological space of kinase inhibitors by introducing the adversarial auto-encoder model (AAE) to improve the generalization of the prediction model. To complete the data space, we constructed Ensemble of PCM-AAE (EPA), an ensemble model that quickly and accurately yields quantitative predictions of binding affinity between any human kinase and inhibitor. In rigorous internal validation, EPA showed excellent performance, consistently outperforming the model trained with the imbalanced set, especially for targets with relatively fewer training data points. Improved prediction accuracy of EPA for external datasets enhances its generalization ability, making it possible to gracefully handle previously unseen kinases and inhibitors. EPA showed promising potential when directly applied to virtual screening and off-target prediction, exhibiting its practicality in hit prediction. Our strategy is expected to facilitate kinase-centric drug development, as well as to solve more challenging prediction problems with insufficient data points.

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.

Analyzing Kinase Similarity in Small Molecule and Protein Structural Space to Explore the Limits of Multi-Target Screening

While selective inhibition is one of the key assets for a small molecule drug, many diseases can only be tackled by simultaneous inhibition of several proteins. An example where achieving selectivity is especially challenging are ligands targeting human kinases. This difficulty arises from the high structural conservation of the kinase ATP binding sites, the area targeted by most inhibitors. We investigated the possibility to identify novel small molecule ligands with pre-defined binding profiles for a series of kinase targets and anti-targets by in silico docking. The candidate ligands originating from these calculations were assayed to determine their experimental binding profiles. Compared to previous studies, the acquired hit rates were low in this specific setup, which aimed at not only selecting multi-target kinase ligands, but also designing out binding to anti-targets. Specifically, only a single profiled substance could be verified as a sub-micromolar, dual-specific EGFR/ErbB2 ligand that indeed avoided its selected anti-target BRAF. We subsequently re-analyzed our target choice and in silico strategy based on these findings, with a particular emphasis on the hit rates that can be expected from a given target combination. To that end, we supplemented the structure-based docking calculations with bioinformatic considerations of binding pocket sequence and structure similarity as well as ligand-centric comparisons of kinases. Taken together, our results provide a multi-faceted picture of how pocket space can determine the success of docking in multi-target drug discovery efforts.

KLIFS: an overhaul after the first 5 years of supporting kinase research

Kinases are a prime target of drug development efforts with >60 drug approvals in the past two decades. Due to the research into this protein family, a wealth of data has been accumulated that keeps on growing. KLIFS-Kinase-Ligand Interaction Fingerprints and Structures-is a structural database focusing on how kinase inhibitors interact with their targets. The aim of KLIFS is to support (structure-based) kinase research through the systematic collection, annotation, and processing of kinase structures. Now, 5 years after releasing the initial KLIFS website, the database has undergone a complete overhaul with a new website, new logo, and new functionalities. In this article, we start by looking back at how KLIFS has been used by the research community, followed by a description of the renewed KLIFS, and conclude with showcasing the functionalities of KLIFS. Major changes include the integration of approved drugs and inhibitors in clinical trials, extension of the coverage to atypical kinases, and a RESTful API for programmatic access. KLIFS is available at the new domain https://klifs.net.

GraphGR: A graph neural network to predict the effect of pharmacotherapy on the cancer cell growth.. [DOI: 10.1101/2020.05.20.107458]

Genomic profiles of cancer cells provide valuable information on genetic alterations in cancer. Several recent studies employed these data to predict the response of cancer cell lines to treatment with drugs. Nonetheless, due to the multifactorial phenotypes and intricate mechanisms of cancer, the accurate prediction of the effect of pharmacotherapy on a specific cell line based on the genetic information alone is problematic. High prediction accuracies reported in the literature likely result from significant overlaps among training, validation, and testing sets, making many predictors inapplicable to new data. To address these issues, we developed GraphGR, a graph neural network with sophisticated attention propagation mechanisms to predict the therapeutic effects of kinase inhibitors across various tumors. Emphasizing on the system-level complexity of cancer, GraphGR integrates multiple heterogeneous data, such as biological networks, genomics, inhibitor profiling, and genedisease associations, into a unified graph structure. In order to construct diverse and information-rich cancer-specific networks, we devised a novel graph reduction protocol based on not only the topological information, but also the biological knowledge. The performance of GraphGR, properly cross-validated at the tissue level, is 0.83 in terms of the area under the receiver operating characteristics, which is notably higher than those measured for other approaches on the same data. Finally, several new predictions are validated against the biomedical literature demonstrating that GraphGR generalizes well to unseen data, i.e. it can predict therapeutic effects across a variety of cancer cell lines and inhibitors. GraphGR is freely available to the academic community at https://github.com/pulimeng/GraphGR.

KinaMetrix: a web resource to investigate kinase conformations and inhibitor space

Protein kinases are among the most explored protein drug targets. Visualization of kinase conformations is critical for understanding structure–function relationship in this family and for developing chemically unique, conformation-specific small molecule drugs. We have developed Kinformation, a random forest classifier that annotates the conformation of over 3500 protein kinase structures in the Protein Data Bank. Kinformation was trained on structural descriptors derived from functionally important motifs to automatically categorize kinases into five major conformations with pharmacological relevance. Here we present KinaMetrix (http://KinaMetrix.com), a web resource enabling researchers to investigate the protein kinase conformational space as well as a subset of kinase inhibitors that exhibit conformational specificity. KinaMetrix allows users to classify uploaded kinase structures, as well as to derive structural descriptors of protein kinases. Uploaded structures can then be compared to atomic structures of other kinases, enabling users to identify kinases that occupy a similar conformational space to their uploaded structure. Finally, KinaMetrix also serves as a repository for both small molecule substructures that are significantly associated with each conformation type, and for homology models of kinases in inactive conformations. We expect KinaMetrix to serve as a resource for researchers studying kinase structural biology or developing conformation-specific kinase inhibitors.

Proteochemometrics - recent developments in bioactivity and selectivity modeling

Proteochemometrics is a machine learning based modeling approach relying on a combination of ligand and protein descriptors. With ongoing developments in machine learning and increases in public data the technique is more frequently applied in early drug discovery, typically in ligand-target binding prediction. Common applications include improvements to single target quantitative structure-activity relationship models, protein selectivity and promiscuity modeling, and large-scale deep learning approaches. The increase in predictive power using proteochemometrics is observed in multi-target bioactivity modeling, opening the door to more extensive studies covering whole protein families. On top of that, with deep learning fueling more complex and larger scale models, proteochemometrics allows faster and higher quality computational models supporting the design, make, test cycle.

Evaluating Polymer Representations via Quantifying Structure-Property Relationships

Machine learning techniques are being applied in quantifying structure-property relationships for a wide variety of materials, where the properly represented materials play key roles. Although algorithms for representation learning are extensively studied, their applications to domain-specific areas, such as polymers, are limited largely due to the lack of benchmark databases. In this work, we investigate different types of polymer representations, including Morgan fingerprint (MF), molecular embedding (ME), and molecular graph (MG), based on the benchmark database from a subset of the well-known web-based polymer databases, PolyInfo. We evaluate the quality of different polymer representations via quantifying the relationships between the representations and polymer properties, including density, melting temperature, and glass transition temperature. Different representation learning schemes for MEs, such as supervised learning, semisupervised learning, and transfer learning, are investigated. In supervised learning, only labeled molecules in our benchmark database are used for representation learning, in semisupervised learning, both labeled and unlabeled molecules in our benchmark database are used, and in transfer learning, molecules from an external database that is different from the benchmark database are used for representation learning. It is found that ME (with the R² of 0.724 in the density case, 0.684 in the melting temperature case, and 0.865 in the glass transition temperature case) outperforms the other representations for structure-property relationship quantification in all cases studied, and MG (with the R² of 0.260 in the density case, -0.149 in the melting temperature case, and 0.711 in the glass transition case) is shown to be much inferior to ME and MF (with the R² of 0.562 in the density case, 0.645 in the melting temperature case, and 0.849 in the glass transition case), likely due to the relatively small volumes of training data available. For MEs, it is found that the similarities of substructure MEs under different learning schemes (e.g., SL, SSL, and TL) are differently estimated, thus leading to different performance scores in structure-property relation quantification. Combinations of MEs show little effect on predictive performance when comparing to the single MEs in the corresponding regression tasks, proving no information gain in mixing MEs.

Using the drug-protein interactome to identify anti-ageing compounds for humans

Advancing age is the dominant risk factor for most of the major killer diseases in developed countries. Hence, ameliorating the effects of ageing may prevent multiple diseases simultaneously. Drugs licensed for human use against specific diseases have proved to be effective in extending lifespan and healthspan in animal models, suggesting that there is scope for drug repurposing in humans. New bioinformatic methods to identify and prioritise potential anti-ageing compounds for humans are therefore of interest. In this study, we first used drug-protein interaction information, to rank 1,147 drugs by their likelihood of targeting ageing-related gene products in humans. Among 19 statistically significant drugs, 6 have already been shown to have pro-longevity properties in animal models (p < 0.001). Using the targets of each drug, we established their association with ageing at multiple levels of biological action including pathways, functions and protein interactions. Finally, combining all the data, we calculated a ranked list of drugs that identified tanespimycin, an inhibitor of HSP-90, as the top-ranked novel anti-ageing candidate. We experimentally validated the pro-longevity effect of tanespimycin through its HSP-90 target in Caenorhabditis elegans.

Human life expectancy is continuing to increase worldwide, as a result of successive improvements in living conditions and medical care. Although this trend is to be celebrated, advancing age is the major risk factor for multiple impairments and chronic diseases. As a result, the later years of life are often spent in poor health and lowered quality of life. However, these effects of ageing are not inevitable, because very long-lived people often suffer rather little ill-health at the end of their lives. Furthermore, laboratory experiments have shown that animals fed with specific drugs can live longer and with fewer age-related diseases than their untreated companions. We therefore need to identify drugs with anti-ageing properties for humans. We have used publically available data and a computer-based approach to search for drugs that affect components and processes known to be important in human ageing. This approach worked, because it was able to re-discover several drugs known to increase lifespan in animal models, plus some new ones, including one that we tested experimentally and validated in this study. These drugs are now a high priority for animal testing and for exploring effects on human ageing.

Drug Discovery Maps, a Machine Learning Model That Visualizes and Predicts Kinome-Inhibitor Interaction Landscapes

The interpretation of high-dimensional structure-activity data sets in drug discovery to predict ligand-protein interaction landscapes is a challenging task. Here we present Drug Discovery Maps (DDM), a machine learning model that maps the activity profile of compounds across an entire protein family, as illustrated here for the kinase family. DDM is based on the t-distributed stochastic neighbor embedding (t-SNE) algorithm to generate a visualization of molecular and biological similarity. DDM maps chemical and target space and predicts the activities of novel kinase inhibitors across the kinome. The model was validated using independent data sets and in a prospective experimental setting, where DDM predicted new inhibitors for FMS-like tyrosine kinase 3 (FLT3), a therapeutic target for the treatment of acute myeloid leukemia. Compounds were resynthesized, yielding highly potent, cellularly active FLT3 inhibitors. Biochemical assays confirmed most of the predicted off-targets. DDM is further unique in that it is completely open-source and available as a ready-to-use executable to facilitate broad and easy adoption.