Issue in running bam_cov.py about basenji HOT 3 CLOSED

Sorry this is overtuned for my setup; I keep an environmental variable pointing to reference genome files. You should be able to dodge this by providing the genome FASTA file with -f. If it's still breaking on the default, just change it to something benign.

I may be overinterpreting your command, but you'll want to be very thoughtful using this program for CLIP-seq modeling. If you just provide large genome blocks like I do for DNase, ATAC, ChIP, and CAGE, you'll be asking the model to learn gene transcript definitions as well as the RNA binding protein sites. I expect that'll be too hard. That's why I focused on CAGE rather than RNA-seq for gene expression. You can provide smaller sequences and make the program behave more like Basset, but I haven't tested that thoroughly. I also haven't tested variable length sequences like you might want for RNA.

from basenji.

i have provided the fasta file using -f but still its breaking, I think i will change the default. Also i am actually looking for RBP binding sites from the CLIP data(as i dont have a CAGE data for this). So my plan was to preprocess the bam, train the sequences and predict motifs. Do you think this approach make sense??

from basenji.

It might work, but you'd likely fit much better models if it better understood the isoform structure of the genes. E.g. I would assemble the training set as RNA sequences annotated with the CLIP coverage or some transformation of it. But in that case, you have variable length sequences, and I don't think Basenji will work out of the box. Maybe you can choose a length that covers most genes and pad the shorter transcripts.

from basenji.