This is a test version of Biostars. For the public version, visit https://www.biostars.org.
COBS Indexing genomes - understand the output

Hi, I am in charge to find presence of specific genes in 600.000 Salmonella's genomes. Some people advised me to use COBS for the indexing So I used it on few genomes just for training

But I don't really understand the output...

I copied a subsequence (55 bp) from one of my genomes, and run COBS to see if it get it. In the output I got 24 (see bellow).

**output:**
SRR18349609 24
SRR18349610 24
SRR18349611 24

Is that mean it got 24 hits on my 55bp query?

And Is it possible to get more information in the output, like e/p value, location etc.. (like blastn)

Thank you for all!!

indexing cobs genomes cobs
• 2,829 views • link
updated by Jianshu_Zhao • Jianshu_Zhao • written by davidmaimoun • davidmaimoun

2 answers

An examlple

$ seqkit  stats t.fasta 
file     format  type  num_seqs  sum_len  min_len  avg_len  max_len
t.fasta  FASTA   DNA          1       60       60       60       60

$ more t.fasta.cobs 
*test   31409
SAMEA18844918   30
SAMN08865341    30

test is the query ID, and31409 is the number of matched records in the COBS index. 30 is the matched k-mers (k=31 by default in COBS), 30 = 60 - 31 + 1 means 100% matched.

For a query of 50 bp, the maximal number of k-mers is 55-31+1 = 25, and SRR18349609 got 24 k-mers matched.

• 0 views • link
written by shenwei356 shenwei356

You may also try KMCP which uses an index structure similar to COBS while with a faster searching speed. While the speedup is not obvious in the 661K dataset cause there are a huge number of similar genomes which results in too many hits for a query, therefore writing results becomes the performance bottleneck.

$ zcat t.fasta.kmcp.tsv.gz  | head -n 4 | column -t 
#query  qLen  qKmers  FPR         hits   target                 chunkIdx  chunks  tLen     kSize  mKmers  qCov    tCov    jacc    queryIdx
test    60    30      2.7536e-05  31406  SAMEA18844918.contigs  0         1       518641   31     30      1.0000  0.0001  0.0001  0
test    60    30      2.7536e-05  31406  SAMN08865341.contigs   0         1       526349   31     30      1.0000  0.0001  0.0001  0
test    60    30      2.7536e-05  31406  SAMEA1462952.contigs   0         1       1199925  31     30      1.0000  0.0000  0.0000  0
• 0 views • link
written by shenwei356 shenwei356

It is looks like great! I'll try it!

Thank you very much

• 0 views • link
written by davidmaimoun • davidmaimoun

Thanks it is very helpful!

And if there is multiple matchs in the same genome (repeated sequences), it will change the output?

• 0 views • link
written by davidmaimoun • davidmaimoun

No, these kinds of methods can't tell the number of sequence copies or sequence locations. They just tell whether the query is contained in the subject, and there are some false positives (see BIGSI/COBS paper).

• 0 views • link
written by shenwei356 shenwei356

I do not think you can get accurate answer using COBS alone because essential it just tells you how many kmers matched. Blastn blastp like alignment-based should be used in the end after some top hits/answers from COBS was found. It was very useful for quick narrow down the target when searching a very large database to only a few candidates that you need to run blastn to confirm.

• 0 views • link
written by Jianshu_Zhao • Jianshu_Zhao

Log in to answer this question.

More posts like this