Could expecto accepts fasta input and return the prdicted tissue-specific expression values?

[WeiCSong]

Hi, thanks for the great tool! I notice that expecto takes a sequence around TSS to predict the tissue-specific expression values and calculate the expression difference between ref and alt sequence. So it is theoretically possible that I input a .fasta sequence that carries multiple variants, and output the corresponding expression values in different tissues and cell types, is that right? Could you tell me whether this is currently applicable, or which part of the code shall I look into and modify to achieve this task? Thanks for your help!

[jzthree]

Do you want to use sequences that are not in the TSSes that we used for training and validation (if it is you can refer to train.py for how to make prediction)? if not, it's possible but not that easy, you will need to construct input features to the linear model (equivalent to ./resources/Xreducedall.2002.npy) using the new fasta file, then train.py has an example of how to make prediction. You can refer to this discussion #9 about how to construct the input feature.

[WeiCSong]

Thanks for the guidance! I'll input the sequence around the same TSS in your paper, but carries multiple variants, so I think I should look into train.py to make a prediction.