Question on the randomer dataset for the RNA inosine_m6A model

[ChaseWLW]

Hi,

I've been using the inosine_m6A model with some RNA004 data. Out of technical curiosity, I would like to ask whether the randomer dataset for training this model includes RNAs that are co-modified with both of these modifications? Or are the sequences either m6A or inosine modified (i.e. not both on the same RNA)?

Thanks!

[marcus1487]

The generation of models are trained from each modified base in isolation. We are looking into adding neighboring bases into training and have done this to a limited extent in DNA models, including multiple 5mC sites in randomers. We are working on ways to scale up this modified neighbors effort into training. Note that we are interested in including all modified bases neighboring each other modified base. For example including A mods (e.g. m6A) next to C mods (e.g. m5C) is on our roadmap to improve the robustness of our models, but is not in the current generation.