Predicting sites of epitranscriptome modifications using unsupervised representation learning based on generative adversarial networks

Abstract

Epitranscriptome is an exciting area that studies different types of modifications in transcripts and the prediction of such modification sites from the transcript sequence is of significant interest. However, the scarcity of positive sites for most modifications imposes critical challenges for training robust algorithms.To circumvent this problem, we propose MR-GAN, a generative adversarial network (GAN) based model, which is trained in an unsupervised fashion on the entire pre-mRNA sequences to learn a low dimensional embedding of transcriptomic sequences. MR-GAN was then applied to extract embeddings of the sequences in a training dataset we created for eight epitranscriptome modifications, including m 6 A, m 1 A, m 1 G, m 2 G, m 5 C, m 5 U, 2′-O-Me, Pseudouridine (Ψ) and Dihydrouridine (D), of which the positive samples are very limited. Prediction models were trained based on the embeddings extracted by MR-GAN. We compared the prediction performance with the one-hot encoding of the training sequences and SRAMP, a state-of-the-art m 6 A site prediction algorithm and demonstrated that the learned embeddings outperform one-hot encoding by a significant margin for up to 15% improvement. Using MR-GAN, we also investigated the sequence motifs for each modification type and uncovered known motifs as well as new motifs not possible with sequences directly. The results demonstrated that transcriptome features extracted using unsupervised learning could lead to high precision for predicting multiple types of epitranscriptome modifications, even when the data size is small and extremely imbalanced.Keywords: N6-methyladenosine (m 6 A), epitranscriptome, RNA modification site prediction, generative adversarial networks (GAN), unsupervised representation learning, methylated RNA immunoprecipitation sequencing (MeRIP-Seq).

Attention-based multi-label neural networks for integrated prediction and interpretation of twelve widely occurring RNA modifications

“…It is thus beneficial and efficient to test the computational framework on multiple RNA modifications simultaneously. Very recently, by taking advantage of the generative adversarial network (GAN), the MR-GAN approach was developed to predict eight RNA modifications 30 . However, some of the modifications supported may be rare modifications, such as m 1 G (only 29 sites), m 2 G (only 59 sites), and D (only 162 sites) 30 , whose wide occurrence in human transcriptome has not yet been confirmed.…”

“…Very recently, by taking advantage of the generative adversarial network (GAN), the MR-GAN approach was developed to predict eight RNA modifications 30 . However, some of the modifications supported may be rare modifications, such as m 1 G (only 29 sites), m 2 G (only 59 sites), and D (only 162 sites) 30 , whose wide occurrence in human transcriptome has not yet been confirmed. Given a large number of negative (non-modifiable) sites of such rare RNA modifications, the sequence-based prediction is likely to produce a substantial proportion of false-positive predictions in practice and should be used with extra caution.…”

Unveiling the Potential Pattern Representation of RNA 5-Methyluridine Modification Sites Through a Novel Feature Fusion Model Leveraging Convolutional Neural Network and Tetranucleotide Composition

“…This novel approach holds promising potential to advance the field of m5U site prediction, representing a noteworthy contribution to the landscape of epitranscriptomics research [12]. Moreover, Diverse cutting-edge classification algorithms, spanning the generalized linear model (GLM) [26], Naive Bayes (NB) [27], random forest (RF) [28], Support Vector Machine (SVM) [29], Particle Swarm Optimization [30], Buffalo-Based Secure Edge-Enabled Computing [31], [32], and neural network-based models [33], were extensively employed to predict m5U sites. In a recent breakthrough, Feng et al [21] developed a machine learning-based computational model, namely iRNA-m5U, to predict m5U sites in RNA.…”