Table 6 Deep learning algorithms in genomics and their original development and applications
From: A review of deep learning applications in human genomics using next-generation sequencing data
ANN Algorithms | Natural Language Processing (NLP) | Feedforward neural network | Convolutional neural network (CNN) | Recurrent neural networks (RNNs) | Bidirectional long short-term memory networks (BLSTMs) | Long short-term memory networks (LSTMs) | Gated recurrent unit (GRU) |
|---|---|---|---|---|---|---|---|
Algorithm Inventor | Applied in dictionary look-up system developed at Birkbeck College, London | Frank Rosenblatt | It was named as “neocognitron “ by Fukushima | Rumelhart, Hinton and Williams | Schuster and Paliwal | Hochreiter and Schmidhuber | Cho et al |
Year of Development | 1948 | 1958 | 1980 | 1986 | 1997 | 1997 | 2014 |
Year of Initial Genomics’ Function | 1996 | 1993 | 2015 | 2005 | 2015 | 2015 | 2017 |
First User in Genomics | Schuler et al | S Eskiizmililer | Alipanahi et al | Maraziotis, Dragomir and Bezerianos | Quang and Xie | Quang and Xie | Angermueller et al |
First Genomic Application | Entrez databases | Karyotyping architecture based on Artificial Neural Networks | DeepBind | Predicting the complicated causative associations between genes from microarray datasets based on recurrent neuro-fuzzy technique | DanQ model | DanQ model | DeepCpG |
Genomic Function Exemplar(s) | Genetic counsellors AI-based chatbots and EPIs prediction | Karyotyping, Prenatal diagnostic for early detection of aneuploidy syndrome | Prediction of variant impacts on expression and disease risk, predicting drug response of tumours from genomic profiles, and pharmacogenomics | Predicting transcription factor binding sites, for Alignment and SNV identification | DNA function predictions and prediction of protein localisation, predict miRNA precursor | Enhancer–promoter interaction (EPI) prediction | Enhancers and methylation states predictions |
Landmark References |