Hi @himanshiarora7

this should be completely possible to use demuxalot for ATAC, though I never tested that.

NH is a tag that is used to detect number of genomic alignments in STAR aligner.

Please sample ~10 random alignments from your bamfile and post them here so I could look at available alignment tags.

Thank you for your response.
I am attaching the sample alignment and the header for my ATAC bam file.

check_atac_bam.txt
header_bam.txt

Hi, I think after recent update I think you can actually use demuxalot for scATAC demuxing, though with some trickery.

You'll need to provide custom callback for parsing reads instead, I think this one should be optimal:

def parse_read(read: AlignedRead) -> Optional[Tuple[float, int]]:
    """
    returns None if read should be ignored.
    Read still can be ignored if it is not in the barcode list
    """
    if read.get_tag("AS") <= len(read.seq) - 6:
        # more than 2 edits
        return None
    if not read.has_tag("UB"):
        # does not have molecule barcode
        return None

    if read.mapq < 20:
        # this one should not be triggered because of NH, but just in case
        return None

    p_misaligned = 0.01  # default value
    fake_ub = random.randint(0, 2 ** 31)
    return p_misaligned, fake_ub

A word of warning: 1) I don't have any scATAC data to test, 2) most important changes in demuxalot are scrnaseq-specific, so you're unlikely to benefit from them