Published October 16, 2026 | Version v2
Dataset Restricted

HMF-m6A: A Hierarchical Multimodal Framework to Intelligently Pinpoint Virus-Induced Host m6A Methylation

  • 1. College of Mathematics, Jilin University, Changchun, Jilin 130012, China
  • 2. State Key Laboratory of Pathogen and Biosecurity, Academy of Military Medical Science, Beijing 100071, China

Description

 It remains computationally challenging to accurately locate host RNA m6A methylation sites induced by viral infections and other stimuli, due to the complexity of local motif contexts. We propose HMF-m6A, a predictive framework to pinpoint RNA m6A methylation sites that hierarchically integrates multimodal features including language model-embedded motifs and their contexts, structural topologies, and physicochemical properties of target RNA sequences, via a learnable sparse gating fusion mechanism. The framework is hierarchically parameterized via an automated motif-directed routing strategy. Additionally, the framework was benchmarked and experimentally validated. HMF-m6A outperformed five contemporary predictors by 8-54% in accuracy and 13-41% in precision. Ablation analyses demonstrated independent contributions of 1%-3% to prediction performance from various features and modules. Moreover, the performance of HMF-m6A was validated in predicting virus-induced host RNA m6A methylation, with an accuracy of 82.7% for HIV, and an average accuracy of 85% across the most frequent motif groups for influenza virus. Particularly, the motif-based hierarchical parameterization was experimentally confirmed in influenza virus-induced m6A methylation, indicating a positive correlation (R^2=0.85) between prediction accuracy and the training dataset size of each motif type. We also developed an interactive web interface and an automated configuration script to facilitate broader application. In summary, the present HMF-m6A markedly improves the accuracy of m6A methylation identification via multimodal feature fusion and fine-grained motif context learning. Its performance has been validated in at least two scenarios of virus-induced host methylation, and the framework will accelerate the profiling of host epigenetic alterations in viral pathogenesis.

Files

Restricted

The record is publicly accessible, but files are restricted. Log in to check if you have access.

Additional details

Dates

Updated
2026-10-28

Software

Repository URL
https://github.com/jiangsy777/HMF-m6A
Programming language
Python