About this Research Topic
A plethora of natural and synthetic chemical compounds are industrially distributed around the world. However, shockingly, the majority of these chemicals have not been comprehensively evaluated for their safety and toxicity due to limitations of standard toxicity testing approaches. Such limitations are due to these methods being performed by in vivo or in vitro analysis for each chemical using experimental analysis or cultured cells, which are often costly and time-consuming processes. Thus, there is a dire need for novel toxicity testing approaches that are more time- and cost-efficient.
High-throughput screening provides a far more efficient alternative to these conventional approaches and has enabled the volume of chemical toxicity data to expand significantly over recent years. However, standard protocols for data analysis are not able to efficiently process such large datasets and highlight a need for novel analytical methods to overcome this challenge. Recently, Deep Learning, a machine learning method using artificial intelligence, has demonstrated the incredible ability to recognize images and extract features from vast input datasets using deep neural networks (DNN). DNNs can extract information from large datasets, automatically, and make high accuracy prediction models utilizing a softmax function that converts an arbitrary numeric output into a “probability value”.
Deep learning has proven to be a useful tool for building Quantitative Structure-Activity Relationship (QSAR) models, defined by quantitative relation build between chemical structure and biological activity used to predict various types of toxicity from large datasets without usage of experimental animals. But, in order to establish a prediction model with high performance, it is first necessary to develop methods to select and extract specific molecular descriptors that provide chemical information contained in the molecules. This approach of using chemical structures as images in the toxicological fields would be a powerful tool to make strong prediction models.
This Research Topic aims to collect original research, review, perspectives, mini review articles highlighting Deep Learning, QSAR and Toxicology.
Keywords: Tox21, Deep Learning, QSAR, Molecular Image, Descriptors
Important Note: All contributions to this Research Topic must be within the scope of the section and journal to which they are submitted, as defined in their mission statements. Frontiers reserves the right to guide an out-of-scope manuscript to a more suitable section or journal at any stage of peer review.