6mAPred-MSFF: A Deep Learning Model for Predicting DNA N6-Methyladenine Sites across Species Based on a Multi-Scale Feature Fusion Mechanism

Abstract

DNA methylation is one of the most extensive epigenetic modifications. DNA N6-methyladenine (6mA) plays a key role in many biology regulation processes. An accurate and reliable genome-wide identification of 6mA sites is crucial for systematically understanding its biological functions. Some machine learning tools can identify 6mA sites, but their limited prediction accuracy and lack of robustness limit their usability in epigenetic studies, which implies the great need of developing new computational methods for this problem. In this paper, we developed a novel computational predictor, namely the 6mAPred-MSFF, which is a deep learning framework based on a multi-scale feature fusion mechanism to identify 6mA sites across different species. In the predictor, we integrate the inverted residual block and multi-scale attention mechanism to build lightweight and deep neural networks. As compared to existing predictors using traditional machine learning, our deep learning framework needs no prior knowledge of 6mA or manually crafted sequence features and sufficiently capture better characteristics of 6mA sites. By benchmarking comparison, our deep learning method outperforms the state-of-the-art methods on the 5-fold cross-validation test on the seven datasets of six species, demonstrating that the proposed 6mAPred-MSFF is more effective and generic. Specifically, our proposed 6mAPred-MSFF gives the sensitivity and specificity of the 5-fold cross-validation on the 6mA-rice-Lv dataset as 97.88% and 94.64%, respectively. Our model trained with the rice data predicts well the 6mA sites of other five species: Arabidopsis thaliana, Fragaria vesca, Rosa chinensis, Homo sapiens, and Drosophila melanogaster with a prediction accuracy 98.51%, 93.02%, and 91.53%, respectively. Moreover, via experimental comparison, we explored performance impact by training and testing our proposed model under different encoding schemes and feature descriptors.

DRSN4mCPred: accurately predicting sites of DNA N4-methylcytosine using deep residual shrinkage network for diagnosis and treatment of gastrointestinal cancer in the precision medicine era

“…The encoding module used four encoding schemes, 1-gram, 2-grams, NAC, and DNC ( 23 ), to represent DNA sequences, resulting in four feature matrices. The features of 1-gram and NAC encoding were combined, as well as the 2-grams and DNC encoding features, and then they were concatenated.…”

“…Next, in the extracting feature module, the matrices of the above two features were fed into the point-wise convolution layer and then supplied to the projection point-wise layer. We extracted more feature information using MS-CAM ( 34 ) network to combine the two matrices features of global and local before they were fused with a Bi-LSTM layer ( 23 ).…”

“…The i4mC-ROSE algorithm ( 19 ) was the first predictive model to predict 4mC sites of Rosaceae genomes and had been used to discriminate 4mC sites of Fragaria vesca ( 20 ) and Rubia chinensis ( 21 ) genomes. The 4mcDeep-CBI ( 22 ) deep learning framework proposed using a 3-convolution neural network (CNN) and Bi-LSTM ( 23 ) to obtain deep information and develop advanced features for discriminating 4mC sites in the DNA sequences of Caenorhabditis elegans ( C. elegans ). The DNC4mC-Deep ( 24 ) utilized several encoding techniques, which included 2Kmer ( 25 ), 3Kmer, binary encoding ( 26 , 27 ), the chemical property and frequency of nucleotides ( 28 ) – along with a CNN and a grid search algorithm to perform 4mC site prediction across cross-species genomes.…”

“…The DRSN4mCPred model used DRSN-CW to eliminate noise-related features and achieved a more accurate feature representation, allowing for better distinguishing of DNA 4mC sites or not. Additionally, inverted residual block, MS-CAM ( 34 ), Bi-LSTM ( 23 ), AFF ( 34 ), and DRSN-CW ( 35 ) were integrated into the prediction model. As a result, it was found that our predictor achieved superior performance in predicting the 4mC sites of different species.…”

MuLan-Methyl - Multiple Transformer-based Language Models for Accurate DNA Methylation Prediction

“…There is a large number of papers that address the problem of identifying methylation sites, however, most of them focus on specific form of modification (10)(11)(12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28)(29), and only a few methods address all three types of methylation mentioned above (30)(31)(32)(33)(34), including iDNA-MS, iDNA-ABT, and iDNA-ABF. Note that the database presented in (31) is now widely used as a benchmark dataset for assessing model performance (21,23,(32)(33)(34).…”