cardioToxCSM: A Web Server for Predicting Cardiotoxicity of Small Molecules

Abstract

The design of novel, safe, and effective drugs to treat human diseases is a challenging venture, with toxicity being one of the main sources of attrition at later stages of development. Failure due to toxicity incurs a significant increase in costs and time to market, with multiple drugs being withdrawn from the market due to their adverse effects. Cardiotoxicity, for instance, was responsible for the failure of drugs such as fenspiride, propoxyphene, and valdecoxib. While significant effort has been dedicated to mitigate this issue by developing computational approaches that aim to identify molecules likely to be toxic, including quantitative structure–activity relationship models and machine learning methods, current approaches present limited performance and interpretability. To overcome these, we propose a new web-based computational method, cardioToxCSM, which can predict six types of cardiac toxicity outcomes, including arrhythmia, cardiac failure, heart block, hERG toxicity, hypertension, and myocardial infarction, efficiently and accurately. cardioToxCSM was developed using the concept of graph-based signatures, molecular descriptors, toxicophore matchings, and molecular fingerprints, leveraging explainable machine learning, and was validated internally via different cross validation schemes and externally via low-redundancy blind sets. The models presented robust performances with areas under ROC curves of up to 0.898 on 5-fold cross-validation, consistent with metrics on blind tests. Additionally, our models provide interpretation of the predictions by identifying whether substructures that are commonly enriched in toxic compounds were present. We believe cardioToxCSM will provide valuable insight into the potential cardiotoxicity of small molecules early on drug screening efforts. The method is made freely available as a web server at .

In Silico Prediction of Human Organ Toxicity via Artificial Intelligence Methods

“…And we believe that our research incorporated large-scale data sets for prediction model construction of human organ level toxicity end points, which is more useful for drug discovery in wide chemical space compared with the above models built by small-scale data sets. For the other toxicity end points, such as cardiotoxicity and developmental toxicity, the AUC values of our models are comparable to the previous study. − …”

Predicting Cardiotoxicity of Molecules Using Attention-Based Graph Neural Networks

“…The data sets used in this study were collected from Iftkhar et al According to their description, a set of 18,803 molecules categorized into six types of cardiotoxicity–cardiac failure, arrhythmia, heart block, myocardial infarction, hypertension, and hERG toxicity–were collected from published studies, including Cai et al, Munawar et al, and Karim et al The data samples in each data set were preprocessed by removing heavy-weight molecules (greater than 1 kDa) since oral drugs usually have smaller weights . To curate these data sets, we followed the curation pipeline originally developed by Fourches et al, with some adjustments described in our previous work .…”

“…Besides, more advanced prediction models, such as pkCSM, DeepHIT, CardioTox net, and cardioToxCSM, were introduced to address this issue with promising outcomes. While pkCSM and cardioToxCSM were developed using graph-based features, , DeepHIT and CardioTox net were combinatory models consisting of three separate neural networks trained with three types of features: molecular descriptors, fingerprints, and graph-based features. , Although using graph-based features for modeling may not capture distinct substructural characteristics of compounds, these lightweight models can serve as high-throughput screening tools. Combinatory or ensemble models can extract substructural characteristics to improve prediction efficiency; however, low-speed processing restricts their application in large-scale screening.…”

“…On the other hand, the lack of experimental evidence for negative samples may cause modeling strategies to face higher risks of false positives. Besides, more advanced prediction models, 23 pkCSM, 24 DeepHIT, 25 CardioTox net, 26 and cardioToxCSM, 27 were introduced to address this issue with promising outcomes. While pkCSM and cardioToxCSM were developed using graphbased features, 24,27 DeepHIT and CardioTox net were combinatory models consisting of three separate neural networks trained with three types of features: molecular descriptors, fingerprints, and graph-based features.…”

Wird KI neue Medikamente gegen Herzkrankheiten hervorbringen?

“…Es gibt eine Reihe von Methoden, bei denen verschiedene molekulare Merkmale genutzt werden, um mithilfe von traditionellen maschinellen Lernverfahren die hERG-Toxizität von niedermolekularen Molekülen vorherzusagen, wie z. B. der cardioToxCSM-Webserver, welcher neben hERG-Toxizität noch 5 weitere Formen an Kardiotoxizität vorhersagen kann [28]. Auch Dockingstrukturen von hERG und möglichen Inhibitoren können bei solchen Vorhersagen als Merkmale miteinbezogen werden [29].…”