LATEST NEWS
Thanks to the development of life-science technologies, a huge amount of data is being produced relative to DNA and RNA sequences in abundance at the individual subject or even individual cell level. Deep learning is transforming the field of many machine-learning applications, such as computer vision and natural language processing, by effectively leveraging the large amount of data and is now emerging as a promising approach for many genomics modelling tasks.
In particular, recurrent neural networks have been used to predict methylation status, alternative splicing, DNA and RNA binding, and protein structure since their framework is suitable to deal with sequential data, such as genomics sequences. On the other hand, convolutional neural networks, which are traditionally applied to image processing and biomedical imaging, have also been used to predict, for example, gene expression from histone modification data, or to classify samples based on gene expression. Autoencoders have been used for protein function prediction and, more recently, for the task of dimensionality reduction and zero imputation in single cell transcriptomic. Indeed, single cell analysis seems to be one of the most prolific field of genomic applications, due to the enormous amount of cells processed in parallel by the current NGS methods, which thus provide sufficient number of samples for deep learning.
Despite the progress in deep learning, its application to genomics data still suffers some drawbacks, such as the difficulty of application when available data are not sufficient to train the model, a situation in which sometimes it is possible to recur to transfer learning. However, the main drawback remains the ability to explain the model outcomes. Indeed, in the field of genomics, beside high accuracy, there is also the need for understanding the underlying reasons for the outcome predictions and for inferring causal relationships among variables and with the outcome, so as to gain biological insight. In this context, explainable artificial intelligence has emerged as a further research direction in genomic applications.
In this special issue, we encourage submission of papers describing novel algorithms and novel applications of deep learning to genomic studies with a focus on their potentialities and challenges and on the direction of explainable artificial intelligence development. Possible topics include, but are not limited to, deep learning algorithms with application on genomics and explainable deep learning for genomics.
Prospective authors are invited to submit their manuscripts electronically after the “open for submissions” date, adhering to the IEEE/ACM Transactions on Computational Biology and Bioinformatics guidelines (https://www.computer.org/csdl/journal/tb/write-for-us/15053). Please submit your papers through the online system (https://mc.manuscriptcentral.com/tcbb-cs) and be sure to select the special-issue name. Manuscripts should not be published or currently submitted for publication elsewhere. Please submit only full papers intended for review, not abstracts, to the ScholarOne portal.
(1) Barbara Di Camillo – University of Padova, Italy, barbara.dicamillo@unipd.it
(2) Giuseppe Nicosia – University of Cambridge, United Kingdom, gn263@cam.ac.uk
