A maximum entropy model for m6A prediction: from human transcripts to influenza A viral segments

M6A, or N6-methyladenosine, is the most common mRNA modification in eukaryotic cells and plays a role in regulating gene expression. Viruses can exploit this cellular mechanism to carry out their own chemical edits, which in turn impacts viral replication and the activation of the innate immune system. Currently, there are limited maps of m6A modifications within viral genomes, making it necessary to employ alternative methods to expedite the study of these processes.

In this research, scientists have developed a new model called MEMAR (Maximum Entropy model for m6A prediction). This model aims to predict the methylation frequency of m6A based on the eight-mers nucleotide motif surrounding the central adenosine. MEMAR is designed to be efficient, straightforward, and adaptable, allowing for the integration of human methylation data and the fine-tuning of the model to account for the specific methylation levels within individual transcripts and the positioning of motifs relative to coding regions.

This flexibility makes MEMAR more effective in predicting m6A modifications in influenza A viral segments compared to existing mammalian models such as DeepSRAMP.

To evaluate the potential for MEMAR to generalize across various datasets, the researchers tested the model using information from human transcripts gathered through GLORI and miCLIP experiments, as well as from HEK293T and HeLa cell lines. Additionally, MEMAR was evaluated based on its performance in differentiating between positive and negative influenza A viral segments, as well as in altered-methylation scenarios.

The researchers also developed a two-state dynamical model that accounts for the interactions between m6A writers and erasers. This model is based on the global methylation percentage and provides insights into how the model's predictive power can be affected by variations in methylation or de-methylation binding constants.