Machine learning approaches to predict drug efficacy and toxicity in oncology

Improved Detection of Drug-Induced Liver Injury by Integrating Predicted In Vivo and In Vitro Data

Drug-induced liver injury (DILI) has been a significant challenge in drug discovery, often leading to clinical trial failures and necessitating drug withdrawals. Over the last decade, the existing suite of in vitro proxy-DILI assays has generally improved at identifying compounds with hepatotoxicity. However, there is considerable interest in enhancing the in silico prediction of DILI because it allows for evaluating large sets of compounds more quickly and cost-effectively, particularly in the early stages of projects. In this study, we aim to study ML models for DILI prediction that first predict nine proxy-DILI labels and then use them as features in addition to chemical structural features to predict DILI. The features include in vitro (e.g., mitochondrial toxicity, bile salt export pump inhibition) data, in vivo (e.g., preclinical rat hepatotoxicity studies) data, pharmacokinetic parameters of maximum concentration, structural fingerprints, and physicochemical parameters. We trained DILI-prediction models on 888 compounds from the DILI data set (composed of DILIst and DILIrank) and tested them on a held-out external test set of 223 compounds from the DILI data set. The best model, DILIPredictor, attained an AUC-ROC of 0.79. This model enabled the detection of the top 25 toxic compounds (2.68 LR+, positive likelihood ratio) compared to models using only structural features (1.65 LR+ score). Using feature interpretation from DILIPredictor, we identified the chemical substructures causing DILI and differentiated cases of DILI caused by compounds in animals but not in humans. For example, DILIPredictor correctly recognized 2-butoxyethanol as nontoxic in humans despite its hepatotoxicity in mice models. Overall, the DILIPredictor model improves the detection of compounds causing DILI with an improved differentiation between animal and human sensitivity and the potential for mechanism evaluation. DILIPredictor required only chemical structures as input for prediction and is publicly available at https://broad.io/DILIPredictor for use via web interface and with all code available for download.

Improved Detection of Drug-Induced Liver Injury by Integrating Predicted in vivo and in vitro Data

Drug-induced liver injury (DILI) has been significant challenge in drug discovery, often leading to clinical trial failures and necessitating drug withdrawals. The existing suite of in vitro proxy-DILI assays is generally effective at identifying compounds with hepatotoxicity. However, there is considerable interest in enhancing in silico prediction of DILI because it allows for the evaluation of large sets of compounds more quickly and cost-effectively, particularly in the early stages of projects. In this study, we aim to study ML models for DILI prediction that first predicts nine proxy-DILI labels and then uses them as features in addition to chemical structural features to predict DILI. The features include in vitro (e.g., mitochondrial toxicity, bile salt export pump inhibition) data, in vivo (e.g., preclinical rat hepatotoxicity studies) data, pharmacokinetic parameters of maximum concentration, structural fingerprints, and physicochemical parameters. We trained DILI-prediction models on 888 compounds from the DILIst dataset and tested on a held-out external test set of 223 compounds from DILIst dataset. The best model, DILIPredictor, attained an AUC-ROC of 0.79. This model enabled the detection of top 25 toxic compounds compared to models using only structural features (2.68 LR+ score). Using feature interpretation from DILIPredictor, we were able to identify the chemical substructures causing DILI as well as differentiate cases DILI is caused by compounds in animals but not in humans. For example, DILIPredictor correctly recognized 2-butoxyethanol as non-toxic in humans despite its hepatotoxicity in mice models. Overall, the DILIPredictor model improves the detection of compounds causing DILI with an improved differentiation between animal and human sensitivity as well as the potential for mechanism evaluation. DILIPredictor is publicly available at https://broad.io/DILIPredictor for use via web interface and with all code available for download and local implementation via https://pypi.org/project/dilipred/.

Deciphering cuproptosis-related signatures in pediatric allergic asthma using integrated scRNA-seq and bulk RNA-seq analysis

OBJECTIVE

Allergic asthma (AA) is common in children. Excess copper is observed in AA patients. It is currently unclear whether copper imbalance can cause cuproptosis in pediatric AA.

METHODS

The datasets about pediatric AA (GSE40732 and GSE40888) were obtained from Gene Expression Omnibus (GEO) database. The expression of cuproptosis-related genes (CRGs) and immune cell infiltration in pediatric AA samples were analyzed. Single-cell RNA sequencing (scRNA-seq) data (GSE193816) were used to evaluate the expression patterns of CRGs in AA. The identification of differentially expressed genes within clusters was conducted using weighted gene co-expression network analysis. Subsequently, disease progression and cuproptosis-related models were screened using random forest (RF), support vector machine (SVM), extreme gradient boosting (XGBoost), and general linear model (GLM) algorithms.

RESULTS

Four CRGs were notably increased in pediatric AA samples. CD4+ T cells, macrophages and mast cells exhibited a lower cuproptosis score in AA samples, indicating that these immune cells may be closely associated with cuproptosis in AA development. Co-expression network of CRGs in AA was constructed. AA samples were divided into two cuprotosis clusters. Following construction of four machine-learning models, SVM model exhibited the highest efficacy of prediction in the testing set (AUC = 0.952). SVM model containing five important variables can be used for prediction of AA.

CONCLUSION

This work provided a machine learning model containing five important variables, which may have good diagnostic efficiency for pediatric AA.

Exploiting Nanotechnology for Drug Delivery: Advancing the Anti-Cancer Effects of Autophagy-Modulating Compounds in Traditional Chinese Medicine

Background

Cancer continues to be a prominent issue in the field of medicine, as demonstrated by recent studies emphasizing the significant role of autophagy in the development of cancer. Traditional Chinese Medicine (TCM) provides a variety of anti-tumor agents capable of regulating autophagy. However, the clinical application of autophagy-modulating compounds derived from TCM is impeded by their restricted water solubility and bioavailability. To overcome this challenge, the utilization of nanotechnology has been suggested as a potential solution. Nonetheless, the current body of literature on nanoparticles delivering TCM-derived autophagy-modulating anti-tumor compounds for cancer treatment is limited, lacking comprehensive summaries and detailed descriptions.

Methods

Up to November 2023, a comprehensive research study was conducted to gather relevant data using a variety of databases, including PubMed, ScienceDirect, Springer Link, Web of Science, and CNKI. The keywords utilized in this investigation included "autophagy", "nanoparticles", "traditional Chinese medicine" and "anticancer".

Results

This review provides a comprehensive analysis of the potential of nanotechnology in overcoming delivery challenges and enhancing the anti-cancer properties of autophagy-modulating compounds in TCM. The evaluation is based on a synthesis of different classes of autophagy-modulating compounds in TCM, their mechanisms of action in cancer treatment, and their potential benefits as reported in various scholarly sources. The findings indicate that nanotechnology shows potential in enhancing the availability of autophagy-modulating agents in TCM, thereby opening up a plethora of potential therapeutic avenues.

Conclusion

Nanotechnology has the potential to enhance the anti-tumor efficacy of autophagy-modulating compounds in traditional TCM, through regulation of autophagy.

Development of a predictive model for nephrotoxicity during tacrolimus treatment using machine learning methods

AIM

When administering tacrolimus, therapeutic drug monitoring is recommended because nephrotoxicity, an adverse event, occurs at supra-therapeutic whole-blood concentrations of tacrolimus. However, some patients exhibit nephrotoxicity even at the recommended concentrations, therefore establishing a therapeutic range of tacrolimus concentration for the individual patient is necessary to avoid nephrotoxicity. This study aimed to develop a model for individualized prediction of nephrotoxicity in patients administered tacrolimus.

METHODS

We collected data, such as laboratory test data at tacrolimus initiation, concomitant drugs and tacrolimus whole-blood concentration, from medical records of patients who received oral tacrolimus. Nephrotoxicity was defined as an increase in serum creatinine levels within 60 days of tacrolimus initiation. We built 13 prediction models based on different machine learning algorithms: logistic regression, support vector machine, gradient-boosting trees, random forest and neural networks. The best performing model was compared with the conventional model, which classifies patients according to the tacrolimus concentration alone.

RESULTS

Data from 163 and 41 patients were used to construct models and evaluate the best performing one, respectively. Most of the patients were diagnosed with inflammatory or autoimmune diseases. The best performing model was built using a support vector machine; it showed a high F2 score of 0.750 and outperformed the conventional model (0.500).

CONCLUSIONS

A machine learning model to predict nephrotoxicity in patients during tacrolimus treatment was developed using tacrolimus whole-blood concentration and other patient data. This model could potentially assist in identifying high-risk patients who require individualized target therapeutic concentrations of tacrolimus prior to treatment initiation to prevent nephrotoxicity.

Leveraging Emerging Technologies to Expand Accessibility and Improve Precision in Rehabilitation and Exercise for People with Disabilities

Physical rehabilitation and exercise training have emerged as promising solutions for improving health, restoring function, and preserving quality of life in populations that face disparate health challenges related to disability. Despite the immense potential for rehabilitation and exercise to help people with disabilities live longer, healthier, and more independent lives, people with disabilities can experience physical, psychosocial, environmental, and economic barriers that limit their ability to participate in rehabilitation, exercise, and other physical activities. Together, these barriers contribute to health inequities in people with disabilities, by disproportionately limiting their ability to participate in health-promoting physical activities, relative to people without disabilities. Therefore, there is great need for research and innovation focusing on the development of strategies to expand accessibility and promote participation in rehabilitation and exercise programs for people with disabilities. Here, we discuss how cutting-edge technologies related to telecommunications, wearables, virtual and augmented reality, artificial intelligence, and cloud computing are providing new opportunities to improve accessibility in rehabilitation and exercise for people with disabilities. In addition, we highlight new frontiers in digital health technology and emerging lines of scientific research that will shape the future of precision care strategies for people with disabilities.

Edge, Fog, and Cloud Against Disease: The Potential of High-Performance Cloud Computing for Pharma Drug Discovery

The high-performance computing (HPC) platform for large-scale drug discovery simulation demands significant investment in speciality hardware, maintenance, resource management, and running costs. The rapid growth in computing hardware has made it possible to provide cost-effective, robust, secure, and scalable alternatives to the on-premise (on-prem) HPC via Cloud, Fog, and Edge computing. It has enabled recent state-of-the-art machine learning (ML) and artificial intelligence (AI)-based tools for drug discovery, such as BERT, BARD, AlphaFold2, and GPT. This chapter attempts to overview types of software architectures for developing scientific software or application with deployment agnostic (on-prem to cloud and hybrid) use cases. Furthermore, the chapter aims to outline how the innovation is disrupting the orthodox mindset of monolithic software running on on-prem HPC and provide the paradigm shift landscape to microservices driven application programming (API) and message parsing interface (MPI)-based scientific computing across the distributed, high-available infrastructure. This is coupled with agile DevOps, and good coding practices, low code and no-code application development frameworks for cost-efficient, secure, automated, and robust scientific application life cycle management.

Repurposing Therapeutic Drugs Complexed to Vanadium in Cancer

Repurposing drugs by uncovering new indications for approved drugs accelerates the process of establishing new treatments and reduces the high costs of drug discovery and development. Metal complexes with clinically approved drugs allow further opportunities in cancer therapy-many vanadium compounds have previously shown antitumor effects, which makes vanadium a suitable metal to complex with therapeutic drugs, potentially improving their efficacy in cancer treatment. In this review, covering the last 25 years of research in the field, we identified non-oncology-approved drugs suitable as ligands to obtain different vanadium complexes. Metformin-decavanadate, vanadium-bisphosphonates, vanadyl(IV) complexes with non-steroidal anti-inflammatory drugs, and cetirizine and imidazole-based oxidovanadium(IV) complexes, each has a parent drug known to have different medicinal properties and therapeutic indications, and all showed potential as novel anticancer treatments. Nevertheless, the precise mechanisms of action for these vanadium compounds against cancer are still not fully understood.

Discovering the Active Ingredients of Medicine and Food Homologous Substances for Inhibiting the Cyclooxygenase-2 Metabolic Pathway by Machine Learning Algorithms

Cyclooxygenase-2 (COX-2) and microsomal prostaglandin E2 synthase (mPGES-1) are two key targets in anti-inflammatory therapy. Medicine and food homology (MFH) substances have both edible and medicinal properties, providing a valuable resource for the development of novel, safe, and efficient COX-2 and mPGES-1 inhibitors. In this study, we collected active ingredients from 503 MFH substances and constructed the first comprehensive MFH database containing 27,319 molecules. Subsequently, we performed Murcko scaffold analysis and K-means clustering to deeply analyze the composition of the constructed database and evaluate its structural diversity. Furthermore, we employed four supervised machine learning algorithms, including support vector machine (SVM), random forest (RF), deep neural networks (DNNs), and eXtreme Gradient Boosting (XGBoost), as well as ensemble learning, to establish 640 classification models and 160 regression models for COX-2 and mPGES-1 inhibitors. Among them, ModelA_ensemble_RF_1 emerged as the optimal classification model for COX-2 inhibitors, achieving predicted Matthews correlation coefficient (MCC) values of 0.802 and 0.603 on the test set and external validation set, respectively. ModelC_RDKIT_SVM_2 was identified as the best regression model based on COX-2 inhibitors, with root mean squared error (RMSE) values of 0.419 and 0.513 on the test set and external validation set, respectively. ModelD_ECFP_SVM_4 stood out as the top classification model for mPGES-1 inhibitors, attaining MCC values of 0.832 and 0.584 on the test set and external validation set, respectively. The optimal regression model for mPGES-1 inhibitors, ModelF_3D_SVM_1, exhibited predictive RMSE values of 0.253 and 0.35 on the test set and external validation set, respectively. Finally, we proposed a ligand-based cascade virtual screening strategy, which integrated the well-performing supervised machine learning models with unsupervised learning: the self-organized map (SOM) and molecular scaffold analysis. Using this virtual screening workflow, we discovered 10 potential COX-2 inhibitors and 15 potential mPGES-1 inhibitors from the MFH database. We further verified candidates by molecular docking, investigated the interaction of the candidate molecules upon binding to COX-2 or mPGES-1. The constructed comprehensive MFH database has laid a solid foundation for the further research and utilization of the MFH substances. The series of well-performing machine learning models can be employed to predict the COX-2 and mPGES-1 inhibitory capabilities of unknown compounds, thereby aiding in the discovery of anti-inflammatory medications. The COX-2 and mPGES-1 potential inhibitor molecules identified through the cascade virtual screening approach provide insights and references for the design of highly effective and safe novel anti-inflammatory drugs.