chris1221 / lanceotron

Command-line interface to the LanceOTron deep learning peak caller

Hi there, I'm Chris.

I am currently a lead scientist at Deep Genomics in Toronto. My work involves using machine learning and AI to identify causal drug targets from human genetics data. I previously completed my DPhil working with Gerton Lunter and Thomas Milne at the University of Oxford. Check out my website for a breakdown of previous projects and repositories of interest.

lanceotron's Issues

Optional argument to threshold peaks on score

This would avoid a post-processing step to filter the peaks based on a score, which can be a hassle

Correct normalisation of ATAC-seq file

Hi,

I have generated two bigwig files synthetically (not as a conversion of bam files) for assemblies hg19 and GRCh38, such that instead of counts I have scores in the interval $[0.0, 1.0]$ (basically, normalising the ATAC-seq signal by the maximum number of counts). Would these be valid inputs to LanceOtron, or do they need to be scaled?

I read in README.md that you advise

We recommend directly converting BAM files to bigwigs with deepTools using the following command:

bamCoverage --bam filename.bam.sorted -o filename.bw --extendReads -bs 1 --normalizeUsing RPKM

Should I try to map my $[0.0, 1.0]$ signals to an RPKM normalisation equivalent, or is the network scaling-agnostic?

Folder needs to exist in order for the results to be written

Would it be possible to create the directory/ parent directories if they do not exist?