Add more epigenetic information to the model

[Lxia626]

Hi,

I'm trying to finetune the model using whole genome-wide methylation sites. I read about your suggestions in #4 , and I'd like to try that. But I have one more question related to the input of the model. I'm wondering if it is possible to add some new channal(s) which contains data from ATAC-seq, ChIP-seq or other epigenetic sequencing methods as the input. If it's possible, could you give some suggestions on this?

[lcamillo]

Hi! The way that CpGPT is setup, the model is unimodal and was designed for DNA methylation. However, in theory it is possible to use any other genomic data as input to the model, as long as there is an associated sequence and an associated value. In the case of methylation, the sequence is +-1000 bp from the CpG site and the value is the methylation proportion. For something like ATAC or ChIP-seq, using something like the TSS +- 1000bp of the genes with the associated log fold change from the control would make sense. Let me know if this is clear and if you have any other questions!