--input-file-type bed expecting 9 columns

[perinom]

Hello,

I'm trying to run IDR (v IDR 2.0.4.2, latest on Conda) with a 6-columns bed file (extended peak summits from MACS2) using --input-file-type bed --rank 5 but it fails with

[...]
File "mydir/snakemake/02/.snakemake/conda/ee8bd795864fb92a3b69fa9e4029ccf7_/lib/python3.7/site-packages/idr/idr.py", line 65, in load_bed
    float(data[6]), float(data[7]), float(data[8])
IndexError: list index out of range

It seems to me IDR is trying to read a 9 column file (like a narrowPeak), and failing. Editing the .bed by adding 3 empty columns and shifting column 5 (with bed scores) to column 7 (narrowPeaks score) works, but it's not the cleanest of the workarounds.

Running IDR with the narrowPeak file from the same MACS call the summits are coming from works, while it fails with the same error with any other 6-column bed I tried.

Am I missing something?

Thanks,
Matteo

[perinom]

yes, indeed.

load_bed() is trying to format bed lines with namedtuple() of fixed length with which isn't flexible for bedfile format.

Peak = namedtuple(
    'Peak', ['chrm', 'strand', 'start', 'stop', 'signal', 'summit', 'signalValue', 'pValue', 'qValue'])

I'm not sure whether IDR later uses signal (which corresponds to the column index given with --rank or signalValue