# Hybrid representation learning for human m6A modifications with chromosome-level generalizability

## TL;DR

This paper introduces new deep learning models for predicting m6A RNA modifications that work well across different chromosomes.

## Contribution

The novel hybrid deep learning models improve m6A prediction with chromosome-level generalizability.

## Key findings

- Hybrid Deep Model achieves the highest accuracy under random splits.
- Hybrid Model shows superior generalization under Leave-One-Chromosome-Out validation.
- Deep global representations may overfit in chromosome-independent settings.

## Abstract

N6-methyladenosine
(m6A) is the most abundant internal modification in eukaryotic mRNA and plays essential roles in post-transcriptional gene regulation. While several deep learning approaches have been proposed to predict m6A sites, most suffer from limited chromosome-level generalizability due to evaluation on randomly split datasets.

In this study, we propose two novel hybrid deep learning models—Hybrid Model and Hybrid Deep Model—that integrate local sequence features (k-mers) and contextual embeddings via convolutional neural networks to improve predictive performance and generalization. We evaluate these models using both a Random-Split strategy and a more biologically realistic Leave-One-Chromosome-Out setting to ensure robustness across genomic regions. Our proposed models outperform the state-of-the-art m6A-TCPred model across all key evaluation metrics. Hybrid Deep Model achieves the highest accuracy under Random-Split, while Hybrid Model demonstrates superior generalization under Leave-One-Chromosome-Out, indicating that deep global representations may overfit in chromosome-independent settings. These findings underscore the importance of rigorous validation strategies and offer insights into designing robust m6A predictors.

Source code and datasets are available at: https://github.com/malikmtahir/LOCO-m6A