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

MEMAR, a maximum entropy model using eight-mer motifs around the central adenosine, predicts m6A methylation frequency on influenza A viral segments better than the mammalian DeepSRAMP method.

• Why it matters: Few m6A viral maps limit mechanistic studies of how viruses hijack host editing and how m6A affects viral replication and innate immune activation.
• What they did: MEMAR is trained on human methylation maps and refined with methylation level per transcript and motif positions relative to coding regions, then tested on GLORI/miCLIP transcripts, HEK293T and HeLa cells, influenza A segments, and altered-methylation experiments.
• The result: A two-state dynamical model between m6A writers and erasers links MEMAR performance to global methylation percentage and interprets altered-condition anomalies via changes in methylation or de-methylation binding constants.