About this Research Topic
Deep learning, also called deep structured learning or hierarchical learning, is an important member of the machine learning method family. It allows computational models that are composed of multiple processing layers to be fed with raw data and automatically learn multiple levels of abstract representations of data for detection and classification.
In recent years, we have witnessed that deep learning has been successfully applied to such diverse fields as image and speech recognition, visual art and natural language processing, drug discovery, bioinformatics, and toxicology. For instance, deep convolutional neural networks have brought about breakthroughs in computer vision and pattern recognition, whereas recurrent neural networks have shed light on sequential data such as text mining and speech applications.
In this Research Topic, we focus on the application of deep learning methods in toxicological and disease diagnosis studies, particularly in qualitative classification and quantitative prediction of toxicity or diseases. In addition, we also seek submissions on how to process high dimensional toxicogenomics or gene expression datasets for de novo reconstruction of gene regulatory or interactive networks. Any toxicological effects or disease phenotypes are the joint action outcome of multiple regulators and effector in a biological system. A systems biology approach is hence warranted. It is our belief that deep learning methods are better suited to identify these causal factors and discover the intricate interactions among them that lead to toxicity or disease. It is also promising to integrate biological and chemical structures into deep learning to discover better data representations and hence, improve the accuracy of disease prediction and toxicity analysis.
Despite great successes, many technical challenges remain, including how to integrate or transform subject-specific knowledge to adapt and improve deep learning algorithms and outcomes such as data preprocessing, model selection (e.g., feedforward, convolutional, or recurrent networks), parametric function approximation (e.g., initialization strategies, activation functions, architecture, and learning techniques), and model regularization and optimization.
We welcome submissions of both Original Research and Review articles that address these existing challenges in the fields of mechanistic and predictive toxicology, drug discover, disease diagnosis, staging, phenotyping and monitoring. We also encourage authors to compare deep learning with conventional machine learning methods in order to highlight significant improvements as a result of deep learning in solving the domain-specific problems.
In recent years, we have witnessed that deep learning has been successfully applied to such diverse fields as image and speech recognition, visual art and natural language processing, drug discovery, bioinformatics, and toxicology. For instance, deep convolutional neural networks have brought about breakthroughs in computer vision and pattern recognition, whereas recurrent neural networks have shed light on sequential data such as text mining and speech applications.
In this Research Topic, we focus on the application of deep learning methods in toxicological and disease diagnosis studies, particularly in qualitative classification and quantitative prediction of toxicity or diseases. In addition, we also seek submissions on how to process high dimensional toxicogenomics or gene expression datasets for de novo reconstruction of gene regulatory or interactive networks. Any toxicological effects or disease phenotypes are the joint action outcome of multiple regulators and effector in a biological system. A systems biology approach is hence warranted. It is our belief that deep learning methods are better suited to identify these causal factors and discover the intricate interactions among them that lead to toxicity or disease. It is also promising to integrate biological and chemical structures into deep learning to discover better data representations and hence, improve the accuracy of disease prediction and toxicity analysis.
Despite great successes, many technical challenges remain, including how to integrate or transform subject-specific knowledge to adapt and improve deep learning algorithms and outcomes such as data preprocessing, model selection (e.g., feedforward, convolutional, or recurrent networks), parametric function approximation (e.g., initialization strategies, activation functions, architecture, and learning techniques), and model regularization and optimization.
We welcome submissions of both Original Research and Review articles that address these existing challenges in the fields of mechanistic and predictive toxicology, drug discover, disease diagnosis, staging, phenotyping and monitoring. We also encourage authors to compare deep learning with conventional machine learning methods in order to highlight significant improvements as a result of deep learning in solving the domain-specific problems.
Keywords: Deep neural networks, Predictive toxicology, Genomics, Systems biology, Disease identification
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.
