Missing matches for toy example

[thekswenson]

Build an index on this toy input:

=> cat test/fake0.fasta
>seq1
acgtaaaacgt
>seq2
acgtataacgt

This step ends without error. I don't know if the log is fishy or not, though. It first says Total input symbols: 22, which is right.

But then the bwtparse step says

Parse file contains 1 words
Computing SA of size 2 over an alphabet of size 2

which sounds strange.

When I use this query

=> cat test/fake0_READS.fa
>query1
taaaa
>query2
acgt

I get this

=> rindex/bin/ri-align count test/fake0.fasta test/fake0_READS.fa
query1    0/5    23
query2    0/4    23
Total time: 0.01604s

which appears to find no matches?

The locate command also gives unexpected output:

=> rindex/bin/ri-align locate test/fake0.fasta test/fake0_READS.fa
@HD    VN:1.6    SO:unknown
@SQ    SN:seq1    LN:11
@SQ    SN:seq2    LN:11
query1    0    seq2    5    255    5M    *    0    0    taaaa    *NH:i:1
query2    0    seq1    12    255    4M    *    0    0    acgt    *NH:i:1
Total time: 0.00009s

[maxrossi91]

Hi @thekswenson,

Thank you very much for reporting this issue. I included your toy datasets in the data/toy folder I was able to reproduce the issue.

• The count issue was resolved in https://github.com/maxrossi91/r-index/commit/87e59bf2fb0124cd95542db880325abfada67e34 . The count issue was due to the fact that the query sequences were lowercase, while the index expects them uppercase. This was not an issue with the locate, and it was resolved by uppercasing the query. The correct output is reported below.

  query1  5/5     1
  query2  4/4     4
  Total time: 0.00016s

• The reporting of bwtparse refer to the PFP components, and it seems to be correct since the PFP has only one phrase.

• The locate had another issue, crashing when the query sequence was found at the beginning of a sequence in the index. This was also resolved. Also fixed in https://github.com/maxrossi91/r-index/commit/87e59bf2fb0124cd95542db880325abfada67e34

• The locate issue you reported seems to be local to the bigbwt construction, since with the sais construction algorithm the result is correct, i.e., when building the index with ri-buildfasta -b sais data/toy/small.fa

  @HD     VN:1.6  SO:unknown
  @SQ     SN:seq1 LN:11
  @SQ     SN:seq2 LN:11
  query1  0       seq1    4       255     5M      *       0       0       TAAAA   *       NH:i:1 
  query2  0       seq2    1       255     4M      *       0       0       acgt    *       NH:i:4 
  query2  256     seq1    8       255     4M      *       0       0       *       *       NH:i:4 
  query2  256     seq1    1       255     4M      *       0       0       *       *       NH:i:4 
  query2  256     seq2    8       255     4M      *       0       0       *       *       NH:i:4 
  Total time: 0.00014s

• This issue with bigbwt is likely to be a corner case for small inputs when the dictionary is composed of only one phrase.

I will continue the investigation.

[thekswenson]

Thanks @maxrossi91 !