Open Hoeze opened 2 years ago
Currently, the APARENT dataloader gets the PolyA sites from the transcript GTF annotation: https://github.com/kipoi/models/blob/ae8cf12cf013d07fd2d6c8ac28a6b82069adac80/APARENT/veff/dataloader.py#L87-L111
@johli Is this a viable implementation of the strategy you explained here? Do you maybe have some example data (vcf file + scores) against which we could compare the Kipoi predictions?
xref: https://github.com/kipoi/models/issues/342 xref: https://github.com/johli/aparent/issues/8
Currently, the APARENT dataloader gets the PolyA sites from the transcript GTF annotation: https://github.com/kipoi/models/blob/ae8cf12cf013d07fd2d6c8ac28a6b82069adac80/APARENT/veff/dataloader.py#L87-L111
@johli Is this a viable implementation of the strategy you explained here? Do you maybe have some example data (vcf file + scores) against which we could compare the Kipoi predictions?
xref: https://github.com/kipoi/models/issues/342 xref: https://github.com/johli/aparent/issues/8