weixie03 / simpleaga Goto Github PK

Right now, parse_bigWigs.py finds all the bigWigs in a directory written into the script as a const variable (by @WeiXie03).
Implement: genomedata-style command line interface with

• -t: explicit track name and path specification
• -s: chromosome sizes file

python parse_bigWigs.py -s <chromosome sizes file> -t <track name> <track file path>  -t <track name> <track file path> -t ...

Currently, using own ugly passing around big fat lists of (processed, cleaned, pruned, etc.) tracks as NumPy arrays through a few functions. Clean up this code by inheriting PyTorch's DataLoader class, which handles random sampling and speeds everything up with multiprocessing too.

💡: Would go well with switch hmmlearn --> pomegranate