Documentation for dedeuplication algorithms used about chromap HOT 4 CLOSED

If you use the atac preset parameters, --remove-pcr-duplicates-at-cell-level will be set. Then the duplicates will be removed at cell level as the 10x "new" method. If you don't have a barcode whitelist, Chromap can still use the uncorrected barcodes to remove duplicates at cell level. Thanks for you suggestions. We will improve the document on this.

from chromap.

Thanks for your quick response - can you confirm duplicates are identified using START/END/BARCODE? How is the primary duplicate chosen? Is this based on the quality score? Can you also comment on how your read trimming is performed? Is this similar to cutadapt?

from chromap.

Thanks for your quick response - can you confirm duplicates are identified using START/END/BARCODE?

If you deduplicate at cell level, then yes. For paired-end reads, the fragment (that contains the two reads) as a whole is considered. So the start and the end of two fragments (and their barcodes) have to be the same to be considered as duplicates.

How is the primary duplicate chosen? Is this based on the quality score?

Mapping with highest MAPQ is chosen as primary for cell-level deduplication. For bulk-level deduplication (same as 10x "old" method), we find the most frequent barcode for the dups, and break ties by the barcode abundance when necessary.

Can you also comment on how your read trimming is performed? Is this similar to cutadapt?

For now, Chromap only does one type of adapter trimming. When two reads are overlapped and the fragments are actually shorter than read length. We found this very common in ATAC-seq. For more details, you can look at the Supplementary Figure 5 in the paper, which is an example and makes things easier to understand.

from chromap.

Thanks so much for all your comments, its really nice to use a well supported tool.

from chromap.