hERG-Att: Self-attention-based deep neural network for predicting hERG blockers

Kim, Hyunho; Nam, Hojung

doi:10.1016/j.compbiolchem.2020.107286

Cited by 36 publications

(22 citation statements)

References 22 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Here, we have collected 15 works from the past five years that employ machine learning-based classification approaches to predict hERG inhibition [38][39][40][41][42][43][44][45][46][47][48][49][50][51]. All of these works apply training datasets of more than 1,000 molecules (and up to tens of thousands in some cases [47,48]), and an overall majority presents two-class (active vs. inactive) classification (with the notable example of the 2015 study of Braga et al, who have introduced a third class of "weak blockers") [38]. Categorizing the molecules into the active and inactive classes is usually done by applying common activity thresholds such as 1 µM, 10 µM or their combination, a comprehensive methodological comparison was presented by Siramshetty et al [44].…”

Section: Herg-mediated Cardiotoxicitymentioning

confidence: 99%

Machine learning models for classification tasks related to drug safety

et al. 2021

View full text Add to dashboard Cite

In this review, we outline the current trends in the field of machine learning-driven classification studies related to ADME (absorption, distribution, metabolism and excretion) and toxicity endpoints from the past six years (2015–2021). The study focuses only on classification models with large datasets (i.e. more than a thousand compounds). A comprehensive literature search and meta-analysis was carried out for nine different targets: hERG-mediated cardiotoxicity, blood–brain barrier penetration, permeability glycoprotein (P-gp) substrate/inhibitor, cytochrome P450 enzyme family, acute oral toxicity, mutagenicity, carcinogenicity, respiratory toxicity and irritation/corrosion. The comparison of the best classification models was targeted to reveal the differences between machine learning algorithms and modeling types, endpoint-specific performances, dataset sizes and the different validation protocols. Based on the evaluation of the data, we can say that tree-based algorithms are (still) dominating the field, with consensus modeling being an increasing trend in drug safety predictions. Although one can already find classification models with great performances to hERG-mediated cardiotoxicity and the isoenzymes of the cytochrome P450 enzyme family, these targets are still central to ADMET-related research efforts. Graphical abstract

show abstract

Section: Herg-mediated Cardiotoxicitymentioning

confidence: 99%

Machine learning models for classification tasks related to drug safety

et al. 2021

View full text Add to dashboard Cite

show abstract

“…By employing a self-attention mechanism, the model learns to not only classify the hERG blockers but also capture the data-specific important substructures from molecular circular fingerprints. The authors confirmed that some of the captured substructures of predicted hERG blockers are related to known hERG-related substructures [175].…”

Section: Toxicitymentioning

confidence: 58%

Artificial Intelligence in Drug Discovery: A Comprehensive Review of Data-driven and Machine Learning Approaches

Kim

Lee

et al. 2020

Biotechnol Bioproc E

Self Cite

View full text Add to dashboard Cite

“…Bioinformaticians are also collectively working with clinicians and scientists to analyze, model, and assess risk using these types of data for individuals or populations for better predictive and preventative healthcare [110]. Machine learning, neural networks, and artificial intelligence (AI) are playing larger roles in precision cardio-oncology investigations [111]. With the shear amount of information and data we are now collecting, it is imperative that we have the mathematical and bioinformatic means to properly gather, normalize, analyze, and interpret these data, for which AI is a logical fit.…”

Section: The Future Of Modeling Using Hipscsmentioning

confidence: 99%

Modeling Precision Cardio-Oncology: Using Human-Induced Pluripotent Stem Cells for Risk Stratification and Prevention

et al. 2021

View full text Add to dashboard Cite

Purpose of Review Cardiovascular toxicity is a leading cause of mortality among cancer survivors and has become increasingly prevalent due to improved cancer survival rates. In this review, we synthesize evidence illustrating how common cancer therapeutic agents, such as anthracyclines, human epidermal growth factors receptors (HER2) monoclonal antibodies, and tyrosine kinase inhibitors (TKIs), have been evaluated in cardiomyocytes (CMs) derived from human-induced pluripotent stem cells (hiPSCs) to understand the underlying mechanisms of cardiovascular toxicity. We place this in the context of precision cardio-oncology, an emerging concept for personalizing the prevention and management of cardiovascular toxicities from cancer therapies, accounting for each individual patient’s unique factors. We outline steps that will need to be addressed by multidisciplinary teams of cardiologists and oncologists in partnership with regulators to implement future applications of hiPSCs in precision cardio-oncology. Recent Findings Current prevention of cardiovascular toxicity involves routine screenings and management of modifiable risk factors for cancer patients, as well as the initiation of cardioprotective medications. Despite recent advancements in precision cardio-oncology, knowledge gaps remain and limit our ability to appropriately predict with precision which patients will develop cardiovascular toxicity. Investigations using patient-specific CMs facilitate pharmacological discovery, mechanistic toxicity studies, and the identification of cardioprotective pathways. Studies with hiPSCs demonstrate that patients with comorbidities have more frequent adverse responses, compared to their counterparts without cardiac disease. Further studies utilizing hiPSC modeling should be considered, to evaluate the impact and mitigation of known cardiovascular risk factors, including blood pressure, body mass index (BMI), smoking status, diabetes, and physical activity in their role in cardiovascular toxicity after cancer therapy. Future real-world applications will depend on understanding the current use of hiPSC modeling in order for oncologists and cardiologists together to inform their potential to improve our clinical collaborative practice in cardio-oncology. Summary When applying such in vitro characterization, it is hypothesized that a safety score can be assigned to each individual to determine who has a greater probability of developing cardiovascular toxicity. Using hiPSCs to create personalized models and ultimately evaluate the cardiovascular toxicity of individuals’ treatments may one day lead to more patient-specific treatment plans in precision cardio-oncology while reducing cardiovascular disease (CVD) morbidity and mortality.

show abstract

hERG-Att: Self-attention-based deep neural network for predicting hERG blockers

Cited by 36 publications

References 22 publications

Machine learning models for classification tasks related to drug safety

Machine learning models for classification tasks related to drug safety

Artificial Intelligence in Drug Discovery: A Comprehensive Review of Data-driven and Machine Learning Approaches

Modeling Precision Cardio-Oncology: Using Human-Induced Pluripotent Stem Cells for Risk Stratification and Prevention

Contact Info

Product

Resources

About