NVIDIA-Genomics-Research / AtacWorks

Deep learning based processing of Atac-seq data
https://clara-parabricks.github.io/AtacWorks/
Other
128 stars 23 forks source link

Error when passing directories for training #232

Closed plantformatics closed 3 years ago

plantformatics commented 3 years ago

Hello,

First off, I want to say thank you for producing this tool. A little bit on the experimental design: I have ATAC-seq data from hundreds of donors, some with very high coverage, some with minimal sequencing depth. I want to train a model on the thoroughly sequenced donors (~25) to de-noise the remaining samples. The heavily sequenced donors are quite diverse and reflect the genetic diversity present in the cohort.

My issue arises when passing AtacWorks train the input data. Specifically, I am getting the following error:

atacworks train: error: argument --cleanpeakfile: expected one argument

I am passing --noisybw, --cleanbw, and --cleanpeakfile directory paths containing the noisy bigwigs, clean bigwigs, and clean narrowPeak files, respectively. I haven't tried using a comma separated list yet. Given the tutorial appendix 3 states that one can pass directories as input arguments, I thought I would I bring this to light so adjustments can either be made to the tutorial or to the code to make specifying input this way possible.

Thanks again for constructing this tool,

Alex

plantformatics commented 3 years ago

Update: Turns out there was a mismatch in variable names (peak vs. peaks) and the peaks folder wasn't being passed to atackworks. Training is running without issue. Sorry for the confusion!