barcode correction of concatenated barcodes does not find any correctable barcodes

[combine-lab]

barcode correction of concatenated barcodes does not find any correctable barcodes

[rob-p]

Hi @kh49,

So this seems related to a strange corner case bug with odd length barcodes an unfiltered permit lists. We should have a fix soon. I'll update here when we do.

Best,
Rob

[rob-p]

Hi @kh49 ,

The odd length barcode issue should be fixed in 0.4.3. Could you try it out and see if it's able to correct more barcodes as expected?

Thanks!
Rob

[kh49]

Hi @rob-p

It works great! With pre-correction, this is the output I'm getting with 0.4.3:

alevin-fry generate-permit-list -d fw -i 765subcr_map -o 765subcr_quant -u ../Combined_Ref/scifiwhitelist.txt
2021-11-11 08:58:44 INFO number of unfiltered bcs read = 283,115,520
2021-11-11 08:58:44 INFO paired : false, ref_count : 376,950, num_chunks : 33,622
2021-11-11 08:58:44 INFO read 2 file-level tags
2021-11-11 08:58:44 INFO read 2 read-level tags
2021-11-11 08:58:44 INFO read 1 alignemnt-level tags
2021-11-11 08:58:44 INFO File-level tag values FileTags { bclen: 29, umilen: 8 }
2021-11-11 08:59:07 INFO observed 168,063,270 reads (160,517,565 orientation consistent) in 33,622 chunks --- max ambiguity read occurs in 2,262 refs
2021-11-11 08:59:07 INFO minimum num reads for barcode pass = 10
2021-11-11 08:59:11 INFO num_passing = 1,939,783
2021-11-11 08:59:13 INFO found 1,939,783 cells with non-trivial number of reads by exact barcode match
2021-11-11 08:59:18 INFO There were 7052344 distinct unmatched barcodes, and 2910710 that can be recovered
2021-11-11 08:59:18 INFO Matching unmatched barcodes to retained barcodes took 5.5256785s
2021-11-11 08:59:18 INFO Of the unmatched barcodes
============
2021-11-11 08:59:18 INFO        3,897,363 had exactly 1 single-edit neighbor in the retained list
2021-11-11 08:59:18 INFO        16,814 had >1 single-edit neighbor in the retained list
2021-11-11 08:59:18 INFO        10,051,177 had no neighbor in the retained list
2021-11-11 08:59:19 INFO total number of distinct corrected barcodes : 2,910,710

Also works without pre-correction, with slightly noisier results as expected:

 alevin-fry generate-permit-list -d fw -i 765sub_map -o 765sub_quant -u ../Combined_Ref/scifiwhitelist.txt
2021-11-11 09:03:10 INFO number of unfiltered bcs read = 283,115,520
2021-11-11 09:03:10 INFO paired : false, ref_count : 376,950, num_chunks : 34,023
2021-11-11 09:03:10 INFO read 2 file-level tags
2021-11-11 09:03:10 INFO read 2 read-level tags
2021-11-11 09:03:10 INFO read 1 alignemnt-level tags
2021-11-11 09:03:10 INFO File-level tag values FileTags { bclen: 29, umilen: 8 }
2021-11-11 09:03:33 INFO observed 170,071,436 reads (162,410,482 orientation consistent) in 34,023 chunks --- max ambiguity read occurs in 2,262 refs
2021-11-11 09:03:33 INFO minimum num reads for barcode pass = 10
2021-11-11 09:03:37 INFO num_passing = 1,897,737
2021-11-11 09:03:39 INFO found 1,897,737 cells with non-trivial number of reads by exact barcode match
2021-11-11 09:03:46 INFO There were 8355292 distinct unmatched barcodes, and 3573212 that can be recovered
2021-11-11 09:03:46 INFO Matching unmatched barcodes to retained barcodes took 6.5993911s
2021-11-11 09:03:46 INFO Of the unmatched barcodes
============
2021-11-11 09:03:46 INFO        6,494,906 had exactly 1 single-edit neighbor in the retained list
2021-11-11 09:03:46 INFO        15,119 had >1 single-edit neighbor in the retained list
2021-11-11 09:03:46 INFO        13,955,301 had no neighbor in the retained list
2021-11-11 09:03:47 INFO total number of distinct corrected barcodes : 3,573,212

Thank you!
Kai

[rob-p]

Awesome! Thanks for getting back so quickly (and for your meticulous issue maintenance — i.e. actually closing the issue when it's resolved).

Best,
Rob