search

kundajelab / bpnet

Toolkit to train base-resolution deep neural networks on functional genomics data and to interpret them
http://bit.ly/bpnet-colab
MIT License
141 stars 33 forks source link

BED file with strands #26

Closed albertjstanley closed 3 years ago

albertjstanley commented 3 years ago

Hi,

I am trying to run the BPNET with Chip-seq data that is stranded. I have peaks specified in a bed file with a column specifying whether the peak belongs to the positive or negative strand. Is there a way to train the BPNET on specifically the stranded data? The bed file used in the BPNET tutorial does not specify the strand for each peak.

Thanks!

Avsecz commented 3 years ago

Hi,

BPNet is trained directly on stranded BigWig files. Peaks are just used to restrict the training set from whole genome to relevant 1kb wide regions. I suggest to either merge the peak files or run the peak caller that uses the alignments from both strands.

Best Ziga

albertjstanley commented 3 years ago

I see, thank you!