Convolutional neural networks for classification of alignments of non-coding RNA sequences

Abstract

MotivationThe convolutional neural network (CNN) has been applied to the classification problem of DNA sequences, with the additional purpose of motif discovery. The training of CNNs with distributed representations of four nucleotides has successfully derived position weight matrices on the learned kernels that corresponded to sequence motifs such as protein-binding sites.ResultsWe propose a novel application of CNNs to classification of pairwise alignments of sequences for accurate clustering of sequences and show the benefits of the CNN method of inputting pairwise alignments for clustering of non-coding RNA (ncRNA) sequences and for motif discovery. Classification of a pairwise alignment of two sequences into positive and negative classes corresponds to the clustering of the input sequences. After we combined the distributed representation of RNA nucleotides with the secondary-structure information specific to ncRNAs and furthermore with mapping profiles of next-generation sequence reads, the training of CNNs for classification of alignments of RNA sequences yielded accurate clustering in terms of ncRNA families and outperformed the existing clustering methods for ncRNA sequences. Several interesting sequence motifs and secondary-structure motifs known for the snoRNA family and specific to microRNA and tRNA families were identified.Availability and implementationThe source code of our CNN software in the deep-learning framework Chainer is available at http://www.dna.bio.keio.ac.jp/cnn/, and the dataset used for performance evaluation in this work is available at the same URL.

A deep learning approach to programmable RNA switches

“…particularly adept at recognizing motifs and long-range interactions in nucleotide sequence data 10,[18][19][20][40][41][42][43][44] . We trained a CNN on a one-hot sequence input, an LSTM on a one-hot sequence input, and a CNN on a two-dimensional, one-hot complementarity map representation input (see "Methods" for complete descriptions of all models).…”

DeepTE: a computational method for de novo classification of transposons with convolutional neural network

“…CNN is an essential model of deep learning, and suitable for identifying sequence profiles, due to its excellent feature extraction capability on high-dimensional data (Kelley et al, 2016;Zeng et al, 2016). The input vector of CNN is primarily based on sequence-derived features, such as the frequency of k-mer occurrence applied in this study and one-hot vector strategy (Aoki and Sakakibara, 2018;Fiannaca et al, 2015;Ghandi et al, 2014;Lee et al, 2011;Nguyen et al, 2016). One apparent advantage of the one-hot vector is to reserve specific position information of each individual nucleotide in sequences.…”

“…One particular deep learning model--Convolutional Neural Network (CNN)--have achieved outstanding performance in image classification, speech recognition, and natural language processing (Krizhevsky et al, 2012;Schmidhuber, 2015). CNN model has also been successfully applied in prediction of unknown sequences profiles or motifs and functional activity discovery, without pre-defining sequence features such as prediction of sequence specificities of DNA-and RNAbinding proteins (Alipanahi et al, 2015), effects of noncoding variants (Zhou and Troyanskaya, 2015), and classification of alignments of noncoding RNA sequences (Alipanahi et al, 2015;Aoki and Sakakibara, 2018;Schmidhuber, 2015;Zeng et al, 2016;Zhou and Troyanskaya, 2015).…”

im6A-TS-CNN: Identifying the N6-Methyladenine Site in Multiple Tissues by Using the Convolutional Neural Network

“…In recent years, Convolutional Neural Network (CNN) has been widely used to solve biological problems. 22 , 27 , 28 The structure of the CNN is shown in Figure 1 . It contains a convolutional layer with 200 filters in which the kernel size is 6.…”