FRAG format

This is an format for the fragment files defined by 10x genomics to store the alignments from single-cell ATAC-seq experiment. It can be regarded as the BEDPE format, containing the chromosome, the leftmost position of aligned fragment, and the rightmost position of the aligned fragment, with two extra columns – the barcode information and the counts of the fragments aligned to the same location with the same barcode. It is usually generated by the cellranger-atac or cellranger-arc pipeline. But this is a fairly straightforward format so one can easily convert other alignment results into this format. An example of the fragment file is like:

chr22   10768839        10769063        AAACGAAAGACTCGGA     2
chr22   11333072        11333249        AAACGAAAGACTCGGA     1
chr22   11363891        11364010        AAACGAAAGACTCGGA     1
...

Fields must be separated by tabs. Plain, gzip and BGZF files are supported; the final record does not require a trailing newline. It receives the same validation and count cap as every other record. Malformed final records raise the usual parser error (for example, fewer than five columns or an end coordinate that does not follow the start).

Leading #, track and browser header lines are skipped, including when the file contains only headers. The Python FragParser API raises ValueError: No valid fragments found in FRAG input if a build has no accepted fragments. An append with no accepted fragments raises the same error when the parser has no prior fragments; otherwise it preserves that parser’s statistics. Statistics belong to the parser instance, rather than the supplied track. Records with a negative start or an empty chromosome are skipped as before.