Predicting chemical toxicity effects based on chemical-chemical interactions

Expanding Predictive Capacities in Toxicology: Insights from Hackathon-Enhanced Data and Model Aggregation

In the realm of predictive toxicology for small molecules, the applicability domain of QSAR models is often limited by the coverage of the chemical space in the training set. Consequently, classical models fail to provide reliable predictions for wide classes of molecules. However, the emergence of innovative data collection methods such as intensive hackathons have promise to quickly expand the available chemical space for model construction. Combined with algorithmic refinement methods, these tools can address the challenges of toxicity prediction, enhancing both the robustness and applicability of the corresponding models. This study aimed to investigate the roles of gradient boosting and strategic data aggregation in enhancing the predictivity ability of models for the toxicity of small organic molecules. We focused on evaluating the impact of incorporating fragment features and expanding the chemical space, facilitated by a comprehensive dataset procured in an open hackathon. We used gradient boosting techniques, accounting for critical features such as the structural fragments or functional groups often associated with manifestations of toxicity.

The Rise of Boron-Containing Compounds: Advancements in Synthesis, Medicinal Chemistry, and Emerging Pharmacology

Boron-containing compounds (BCC) have emerged as important pharmacophores. To date, five BCC drugs (including boronic acids and boroles) have been approved by the FDA for the treatment of cancer, infections, and atopic dermatitis, while some natural BCC are included in dietary supplements. Boron's Lewis acidity facilitates a mechanism of action via formation of reversible covalent bonds within the active site of target proteins. Boron has also been employed in the development of fluorophores, such as BODIPY for imaging, and in carboranes that are potential neutron capture therapy agents as well as novel agents in diagnostics and therapy. The utility of natural and synthetic BCC has become multifaceted, and the breadth of their applications continues to expand. This review covers the many uses and targets of boron in medicinal chemistry.

Accurate clinical toxicity prediction using multi-task deep neural nets and contrastive molecular explanations

Explainable machine learning for molecular toxicity prediction is a promising approach for efficient drug development and chemical safety. A predictive ML model of toxicity can reduce experimental cost and time while mitigating ethical concerns by significantly reducing animal and clinical testing. Herein, we use a deep learning framework for simultaneously modeling in vitro, in vivo, and clinical toxicity data. Two different molecular input representations are used; Morgan fingerprints and pre-trained SMILES embeddings. A multi-task deep learning model accurately predicts toxicity for all endpoints, including clinical, as indicated by the area under the Receiver Operator Characteristic curve and balanced accuracy. In particular, pre-trained molecular SMILES embeddings as input to the multi-task model improved clinical toxicity predictions compared to existing models in MoleculeNet benchmark. Additionally, our multitask approach is comprehensive in the sense that it is comparable to state-of-the-art approaches for specific endpoints in in vitro, in vivo and clinical platforms. Through both the multi-task model and transfer learning, we were able to indicate the minimal need of in vivo data for clinical toxicity predictions. To provide confidence and explain the model's predictions, we adapt a post-hoc contrastive explanation method that returns pertinent positive and negative features, which correspond well to known mutagenic and reactive toxicophores, such as unsubstituted bonded heteroatoms, aromatic amines, and Michael receptors. Furthermore, toxicophore recovery by pertinent feature analysis captures more of the in vitro (53%) and in vivo (56%), rather than of the clinical (8%), endpoints, and indeed uncovers a preference in known toxicophore data towards in vitro and in vivo experimental data. To our knowledge, this is the first contrastive explanation, using both present and absent substructures, for predictions of clinical and in vivo molecular toxicity.

Deep learning integration of molecular and interactome data for protein-compound interaction prediction

Motivation

Virtual screening, which can computationally predict the presence or absence of protein–compound interactions, has attracted attention as a large-scale, low-cost, and short-term search method for seed compounds. Existing machine learning methods for predicting protein–compound interactions are largely divided into those based on molecular structure data and those based on network data. The former utilize information on proteins and compounds, such as amino acid sequences and chemical structures; the latter rely on interaction network data, such as protein–protein interactions and compound–compound interactions. However, there have been few attempts to combine both types of data in molecular information and interaction networks.

Results

We developed a deep learning-based method that integrates protein features, compound features, and multiple types of interactome data to predict protein–compound interactions. We designed three benchmark datasets with different difficulties and applied them to evaluate the prediction method. The performance evaluations show that our deep learning framework for integrating molecular structure data and interactome data outperforms state-of-the-art machine learning methods for protein–compound interaction prediction tasks. The performance improvement is statistically significant according to the Wilcoxon signed-rank test. This finding reveals that the multi-interactome data captures perspectives other than amino acid sequence homology and chemical structure similarity and that both types of data synergistically improve the prediction accuracy. Furthermore, experiments on the three benchmark datasets show that our method is more robust than existing methods in accurately predicting interactions between proteins and compounds that are unseen in training samples.

Identifying and Prioritizing Chemicals with Uncertain Burden of Exposure: Opportunities for Biomonitoring and Health-Related Research

BACKGROUND

The National Institutes of Health's Environmental influences on Child Health Outcomes (ECHO) initiative aims to understand the impact of environmental factors on childhood disease. Over 40,000 chemicals are approved for commercial use. The challenge is to prioritize chemicals for biomonitoring that may present health risk concerns.

OBJECTIVES

Our aim was to prioritize chemicals that may elicit child health effects of interest to ECHO but that have not been biomonitored nationwide and to identify gaps needing additional research.

METHODS

We searched databases and the literature for chemicals in environmental media and in consumer products that were potentially toxic. We selected chemicals that were not measured in the National Health and Nutrition Examination Survey. From over 700 chemicals, we chose 155 chemicals and created eight chemical panels. For each chemical, we compiled biomonitoring and toxicity data, U.S. Environmental Protection Agency exposure predictions, and annual production usage. We also applied predictive modeling to estimate toxicity. Using these data, we recommended chemicals either for biomonitoring, to be deferred pending additional data, or as low priority for biomonitoring.

RESULTS

For the 155 chemicals, 97 were measured in food or water, 67 in air or house dust, and 52 in biospecimens. We found in vivo endocrine, developmental, reproductive, and neurotoxic effects for 61, 74, 47, and 32 chemicals, respectively. Eighty-six had data from high-throughput in vitro assays. Positive results for endocrine, developmental, neurotoxicity, and obesity were observed for 32, 11, 35, and 60 chemicals, respectively. Predictive modeling results suggested 90% are toxicants. Biomarkers were reported for 76 chemicals. Thirty-six were recommended for biomonitoring, 108 deferred pending additional research, and 11 as low priority for biomonitoring.

DISCUSSION

The 108 deferred chemicals included those lacking biomonitoring methods or toxicity data, representing an opportunity for future research. Our evaluation was, in general, limited by the large number of unmeasured or untested chemicals. https://doi.org/10.1289/EHP5133.

End-to-End Representation Learning for Chemical-Chemical Interaction Prediction

Chemical-chemical interaction (CCI) plays a major role in predicting candidate drugs, toxicities, therapeutic effects, and biological functions. CCI is typically inferred from a variety of information; however, CCI has yet not been predicted using a learning-based approach. In other drug analyses, deep learning has been actively used in recent years. However, in most cases, deep learning has been used only for classification even though it has feature extraction capabilities. Thus, in this paper, we propose an end-to-end representation learning method for CCI, named DeepCCI, which includes feature extraction and a learning-based approach. Our proposed architecture is based on the Siamese network. Hidden representations are extracted from a simplified molecular input line entry system (SMILES), which is a string notation representing the chemical structure using weight-shared convolutional neural networks. Subsequently, L1 element-wise distances between the two extracted hidden representations are measured. The performance of DeepCCI is compared with those of 12 fingerprint-method combinations. The proposed DeepCCI shows the best performance in most of the evaluation metrics used. In addition, DeepCCI was experimentally validated to guarantee the commutative property. The automatically extracted features can alleviate the efforts required for manual feature engineering and improve prediction performance.

In silico toxicology: comprehensive benchmarking of multi-label classification methods applied to chemical toxicity data

One goal of toxicity testing, among others, is identifying harmful effects of chemicals. Given the high demand for toxicity tests, it is necessary to conduct these tests for multiple toxicity endpoints for the same compound. Current computational toxicology methods aim at developing models mainly to predict a single toxicity endpoint. When chemicals cause several toxicity effects, one model is generated to predict toxicity for each endpoint, which can be labor and computationally intensive when the number of toxicity endpoints is large. Additionally, this approach does not take into consideration possible correlation between the endpoints. Therefore, there has been a recent shift in computational toxicity studies toward generating predictive models able to predict several toxicity endpoints by utilizing correlations between these endpoints. Applying such correlations jointly with compounds' features may improve model's performance and reduce the number of required models. This can be achieved through multi‐label classification methods. These methods have not undergone comprehensive benchmarking in the domain of predictive toxicology. Therefore, we performed extensive benchmarking and analysis of over 19,000 multi‐label classification models generated using combinations of the state‐of‐the‐art methods. The methods have been evaluated from different perspectives using various metrics to assess their effectiveness. We were able to illustrate variability in the performance of the methods under several conditions. This review will help researchers to select the most suitable method for the problem at hand and provide a baseline for evaluating new approaches. Based on this analysis, we provided recommendations for potential future directions in this area.

This article is categorized under: 1

Computer and Information Science > Chemoinformatics

2

Computer and Information Science > Computer Algorithms and Programming

Current data regarding the structure-toxicity relationship of boron-containing compounds

Boron is ubiquitous in nature, being an essential element of diverse cells. As a result, humans have had contact with boron containing compounds (BCCs) for a long time. During the 20th century, BCCs were developed as antiseptics, antibiotics, cosmetics and insecticides. Boric acid was freely used in the nosocomial environment as an antiseptic and sedative salt, leading to the death of patients and an important discovery about its critical toxicology for humans. Since then the many toxicological studies done in relation to BCCs have helped to establish the proper limits of their use. During the last 15 years, there has been a boom of research on the design and use of new, potent and efficient boron containing drugs, finding that the addition of boron to some known drugs increases their affinity and selectivity. This mini-review summarizes two aspects of BCCs: toxicological data found with experimental models, and the scarce but increasing data about the structure-activity relationship for toxicity and therapeutic use. As is the case with boron-free compounds, the biological activity of BCCs is related to their chemical structure. We discuss the use of new technology to discover potent and efficient BCCs for medicinal therapy by avoiding toxic effects.

Analysis of the chemical toxicity effects using the enrichment of Gene Ontology terms and KEGG pathways

BACKGROUND

Chemical toxicity is one of the major barriers for designing and detecting new chemical entities during drug discovery. Unexpected toxicity of an approved drug may lead to withdrawal from the market and significant loss of the associated costs. Better understanding of the mechanisms underlying various toxicity effects can help eliminate unqualified candidate drugs in early stages, allowing researchers to focus their attention on other more viable candidates.

METHODS

In this study, we aimed to understand the mechanisms underlying several toxicity effects using Gene Ontology (GO) terms and KEGG pathways. GO term and KEGG pathway enrichment theories were adopted to encode each chemical, and the minimum redundancy maximum relevance (mRMR) was used to analyze the GO terms and the KEGG pathways. Based on the feature list obtained by the mRMR method, the most related GO terms and KEGG pathways were extracted.

RESULTS

Some important GO terms and KEGG pathways were uncovered, which were concluded to be significant for determining chemical toxicity effects.

CONCLUSIONS

Several GO terms and KEGG pathways are highly related to all investigated toxicity effects, while some are specific to a certain toxicity effect.

GENERAL SIGNIFICANCE

The findings in this study have the potential to further our understanding of different chemical toxicity mechanisms and to assist scientists in developing new chemical toxicity prediction algorithms. This article is part of a Special Issue entitled "System Genetics" Guest Editor: Dr. Yudong Cai and Dr. Tao Huang.

Identification of Chemical Toxicity Using Ontology Information of Chemicals

With the advance of the combinatorial chemistry, a large number of synthetic compounds have surged. However, we have limited knowledge about them. On the other hand, the speed of designing new drugs is very slow. One of the key causes is the unacceptable toxicities of chemicals. If one can correctly identify the toxicity of chemicals, the unsuitable chemicals can be discarded in early stage, thereby accelerating the study of new drugs and reducing the R&D costs. In this study, a new prediction method was built for identification of chemical toxicities, which was based on ontology information of chemicals. By comparing to a previous method, our method is quite effective. We hope that the proposed method may give new insights to study chemical toxicity and other attributes of chemicals.

Prediction of drug indications based on chemical interactions and chemical similarities

Discovering potential indications of novel or approved drugs is a key step in drug development. Previous computational approaches could be categorized into disease-centric and drug-centric based on the starting point of the issues or small-scaled application and large-scale application according to the diversity of the datasets. Here, a classifier has been constructed to predict the indications of a drug based on the assumption that interactive/associated drugs or drugs with similar structures are more likely to target the same diseases using a large drug indication dataset. To examine the classifier, it was conducted on a dataset with 1,573 drugs retrieved from Comprehensive Medicinal Chemistry database for five times, evaluated by 5-fold cross-validation, yielding five 1st order prediction accuracies that were all approximately 51.48%. Meanwhile, the model yielded an accuracy rate of 50.00% for the 1st order prediction by independent test on a dataset with 32 other drugs in which drug repositioning has been confirmed. Interestingly, some clinically repurposed drug indications that were not included in the datasets are successfully identified by our method. These results suggest that our method may become a useful tool to associate novel molecules with new indications or alternative indications with existing drugs.

Integration of bioinformatics to biodegradation

Bioinformatics and biodegradation are two primary scientific fields in applied microbiology and biotechnology. The present review describes development of various bioinformatics tools that may be applied in the field of biodegradation. Several databases, including the University of Minnesota Biocatalysis/Biodegradation database (UM-BBD), a database of biodegradative oxygenases (OxDBase), Biodegradation Network-Molecular Biology Database (Bionemo) MetaCyc, and BioCyc have been developed to enable access to information related to biochemistry and genetics of microbial degradation. In addition, several bioinformatics tools for predicting toxicity and biodegradation of chemicals have been developed. Furthermore, the whole genomes of several potential degrading bacteria have been sequenced and annotated using bioinformatics tools.

Beyond new chemical entities: advancing drug development based on functional versatility of antibodies

Antibody-type agents (i.e., antibodies and derivatives thereof) may be produced as clinically valuable antidotes, which conceivably could be developed in tandem with prospective new pharmaceutical products so as to render the risks of clinical trials more acceptable from a regulatory standpoint. Yet, this is but a relatively narrow view of the full potential utility associated with antibody-type agents, the significance of which is appreciated upon reconsidering key aspects of early modern biomedical research (notably major contributions thereto by Nobel Laureate Paul Ehrlich) in light of much more recent advances (e.g., application of immunity-oriented approaches to diseases in general, epitope-specific targeting, abzyme-mediated catalysis, antibody-mediated sustained-release buffering of unbound-ligand concentrations, and enhanced thermal and metabolic stability of deuterated chemical species via the kinetic isotope effect), as conditioned by health-related concerns (e.g., current and anticipated epidemiologic transitions vis-a-vis environmental changes) especially with regard to sustainable development (e.g., emphasizing more efficient resource utilization toward increased global resilience based on greater independence from high-maintenance technological infrastructure). The broader view that thus emerges highlights the urgent need to rebalance the health-research agenda, which presently reflect an overemphasis on small-molecule candidate-drug discovery, in order to advance health based on a comprehensive fundamental synthesis of immunity and pharmacology.

Prediction of cancer drugs by chemical-chemical interactions

Cancer, which is a leading cause of death worldwide, places a big burden on health-care system. In this study, an order-prediction model was built to predict a series of cancer drug indications based on chemical-chemical interactions. According to the confidence scores of their interactions, the order from the most likely cancer to the least one was obtained for each query drug. The 1(st) order prediction accuracy of the training dataset was 55.93%, evaluated by Jackknife test, while it was 55.56% and 59.09% on a validation test dataset and an independent test dataset, respectively. The proposed method outperformed a popular method based on molecular descriptors. Moreover, it was verified that some drugs were effective to the 'wrong' predicted indications, indicating that some 'wrong' drug indications were actually correct indications. Encouraged by the promising results, the method may become a useful tool to the prediction of drugs indications.

Inferring drug-disease associations from integration of chemical, genomic and phenotype data using network propagation

BACKGROUND

During the last few years, the knowledge of drug, disease phenotype and protein has been rapidly accumulated and more and more scientists have been drawn the attention to inferring drug-disease associations by computational method. Development of an integrated approach for systematic discovering drug-disease associations by those informational data is an important issue.

METHODS

We combine three different networks of drug, genomic and disease phenotype and assign the weights to the edges from available experimental data and knowledge. Given a specific disease, we use our network propagation approach to infer the drug-disease associations.

RESULTS

We apply prostate cancer and colorectal cancer as our test data. We use the manually curated drug-disease associations from comparative toxicogenomics database to be our benchmark. The ranked results show that our proposed method obtains higher specificity and sensitivity and clearly outperforms previous methods. Our result also show that our method with off-targets information gets higher performance than that with only primary drug targets in both test data.

CONCLUSIONS

We clearly demonstrate the feasibility and benefits of using network-based analyses of chemical, genomic and phenotype data to reveal drug-disease associations. The potential associations inferred by our method provide new perspectives for toxicogenomics and drug reposition evaluation.

Advanced systems biology methods in drug discovery and translational biomedicine

Systems biology is in an exponential development stage in recent years and has been widely utilized in biomedicine to better understand the molecular basis of human disease and the mechanism of drug action. Here, we discuss the fundamental concept of systems biology and its two computational methods that have been commonly used, that is, network analysis and dynamical modeling. The applications of systems biology in elucidating human disease are highlighted, consisting of human disease networks, treatment response prediction, investigation of disease mechanisms, and disease-associated gene prediction. In addition, important advances in drug discovery, to which systems biology makes significant contributions, are discussed, including drug-target networks, prediction of drug-target interactions, investigation of drug adverse effects, drug repositioning, and drug combination prediction. The systems biology methods and applications covered in this review provide a framework for addressing disease mechanism and approaching drug discovery, which will facilitate the translation of research findings into clinical benefits such as novel biomarkers and promising therapies.

Identifying chemicals with potential therapy of HIV based on protein-protein and protein-chemical interaction network

Acquired immune deficiency syndrome (AIDS) is a severe infectious disease that causes a large number of deaths every year. Traditional anti-AIDS drugs directly targeting the HIV-1 encoded enzymes including reverse transcriptase (RT), protease (PR) and integrase (IN) usually suffer from drug resistance after a period of treatment and serious side effects. In recent years, the emergence of numerous useful information of protein-protein interactions (PPI) in the HIV life cycle and related inhibitors makes PPI a new way for antiviral drug intervention. In this study, we identified 26 core human proteins involved in PPI between HIV-1 and host, that have great potential for HIV therapy. In addition, 280 chemicals that interact with three HIV drugs targeting human proteins can also interact with these 26 core proteins. All these indicate that our method as presented in this paper is quite promising. The method may become a useful tool, or at least plays a complementary role to the existing method, for identifying novel anti-HIV drugs.