Deep learning in bioinformatics and biomedicine

References

Chat With PDF

Cite

Share

Abstract

Deep learning is a subfield of machine learning that considers computational models with multiple processing layers [1, 3, 6]. At the core of all deep learning approaches lies ‘representation learning’: the models automatically learn a representation of the input data without the explicit guidance of a domain expert. Low-level features (such as edges in image data) are moved forward to the next layer where higher level, more abstract features (such as shapes) are extracted. This feature extraction is based on nonlinear functions in the processing units. Thereby, deep learning is capable of discovering the intrinsic hierarchies in the training data that can be exploited for a variety of analytical tasks. Most deep learning methods are designed for supervised classification, that is tasks for which an input– output mapping has to be learned from labeled training data. The success and failure of predictive models crucially depend on how well the hierarchical structures can be captured. ‘Deep neural networks’, such as multilayer feed-forward perceptrons, convolutional neural networks and recurrent neural networks, are particularly good at this task. Deep learning requires truly large amounts of training data, which became widely available with the dawn of the big data era. The advances in experimental capabilities have led to increasing amounts of complex data in the life sciences, too. The advent of high-throughput technologies brought a paradigm shift from traditionally hypothesis-driven to data-driven research. Studies in biology and medicine rests on ever-growing volumes of complex data that characterize phenomena across a wide range of physical and organizational scales, from molecules to environments [10, 11]. More data hold the promise of a better understanding of the mechanisms underlying many biological structures and (patho-)physiological processes [7], which might ultimately lead to improved therapies for patients. As many problems in the life sciences require the decoding of complex interactions between entities (e.g. genes, proteins), and given the wealth of multimodal data, deep learning methods are believed to have a transformative impact on biomedicine [2]. The aim of this special issue is to provide the readers with a set of reviews that describe the latest concepts, innovations, approaches and technologies in the area of deep learning in bioinformatics, computational biology and systems medicine. In their article titled ‘Biological network analysis with deep learning’, Muzio et al. review the state of the art of deep learning methods for the analysis of graph data. Biological data are often represented in the form of graphs, such as protein interaction networks and gene regulatory networks. However,