Related Experiment Video
Updated: Sep 24, 2025
08:50
Methylated RNA Immunoprecipitation Assay to Study m5C Modification in Arabidopsis
Published on: May 14, 2020
6.8K
Deepm5C: A deep-learning-based hybrid framework for identifying human RNA N5-methylcytosine sites using a stacking
Md Mehedi Hasan1, Sho Tsukiyama2, Jae Youl Cho3
1Tulane Center for Biomedical Informatics and Genomics, Division of Biomedical Informatics and Genomics, John W. Deming Department of Medicine, School of Medicine, Tulane University, New Orleans, LA 70112, USA.
Summary
Deepm5C accurately identifies N5-methylcytosine (m5C) RNA sites using a novel deep learning approach. This bioinformatics tool enhances genome-wide m5C detection, improving understanding of cellular processes and disease.
Area of Science:
- Epigenetics
- Bioinformatics
- Genomics
Background:
- N5-methylcytosine (m5C) is a prevalent RNA epigenetic modification crucial for cellular functions and disease.
- Existing computational methods for m5C identification are limited by small training datasets, hindering genome-wide applications.
- Accurate identification of m5C sites is vital for understanding cellular mechanisms and disease pathogenesis.
Purpose of the Study:
- To develop Deepm5C, a novel bioinformatics method for accurate genome-wide identification of RNA m5C sites in the human genome.
- To overcome limitations of existing methods by utilizing a large, novel benchmarking dataset and advanced computational strategies.
- To provide a robust tool for researchers studying m5C modifications and their roles in biological processes.
Main Methods:
- Construction of a novel benchmarking dataset for RNA m5C site identification.
- Integration of conventional feature encoding algorithms with word-embedding approaches.
- Development and comparison of 32 baseline models using deep learning and conventional classifiers.
- Implementation of a stacking strategy with a 1D convolutional neural network on optimal baseline models.
Main Results:
- Deepm5C achieved high performance metrics, including a Matthews correlation coefficient of 0.697 and accuracy of 0.855 in cross-validation.
- Independent testing yielded comparable results with a Matthews correlation coefficient of 0.691 and accuracy of 0.852.
- Deepm5C demonstrated superior accuracy and stability compared to baseline models and existing predictors.
Conclusions:
- Deepm5C offers a significant advancement in identifying RNA m5C sites, outperforming current methods.
- The hybrid deep learning framework provides a more accurate and stable predictor for m5C identification.
- Deepm5C is expected to aid researchers in identifying potential m5C sites and generating new biological hypotheses.
