Effect of sequence variants on RBP motifs

[gouthamatla]

Hi All,

Thanks for this very useful tool.

I am interested to use the DeepRiPe models to identify RBP binding sites that were disrupted by sequence variants i.e sQTLs. From the Notebook Mutation.ipynb I understood that we need to know a priory what RBP is binding on a given sequence as its shown with ELAVL1 as an example (cell 4 of the notebook) to know if it effects WT/MT sequence.

As mentioned in the paper, The naive approach to associate SNVs with alteration of RBP binding sites is to find mutations that have been mapped to RBP targets obtained from CLIP experiments. However, the resolution of peaks is typically not sufficient to conclude that any mutation will alter RBP binding. Furthermore, CLIP experiment of one cell type or tissue may miss targets in other cell types due to tissue-specific gene expression. Here, our interpretable model provides an opportunity for the identification of mutations that potentially alter the RBP binding sites.

Is there a way to generalize this as an in-silico mutagenesis and evaluate, for each sQTL, which RBP site or which atribution map is potentially disrupted without having the RBP data for my tissue of interest ? I guess I need to take +/- 50bp sequence around my variant of interested and run it for all RBPs and check which one is affected ? Do you provide any score for motif disruption ?

Thanks,
Goutham A