Best match Blast

Question

[aka]

Hi!!

I have results from a tblastx but I would like to keep aonly the best result.

I have this extract:

Query= AT3G23790.1 | Symbols: AAE16 | AMP-dependent synthetase and ligase
family protein | chr3:8575268-8581001 FORWARD LENGTH=2169

Length=2169
                                                                      Score     E
Sequences producing significant alignments:                          (Bits)  Value  N

Qrob_P0041780.2 528                                                   193     5e-49  1
Qrob_P0179090.2 2106                                                  71.2    1e-38  5
Qrob_P0041760.2 606                                                   150     6e-36  1
Qrob_P0279400.2 1896                                                  72.1    4e-34  4
Qrob_P0041770.2 387                                                   84.5    1e-30  2
Qrob_P0278330.2 1632                                                  85.4    2e-25  3
Qrob_P0388410.2 1629                                                  78.5    3e-21  3

What I understand is that I have the best result with the fisrt row Qrob_P0041780.2 528, because I have an high Score and a good E_value but is it the good things? Or it is not the first row?

Some query have an E-value of 0.0, I read some comments and it means that it's a really best score, is it right?

Thank you for your help.

Aka

[Nitin Narwade]

Hmmm

I would say only considering the E-value cut-off is not a very good measure. Rather, I would consider multiple criteria such as Percent Identity, E-value, query coverage, bits score, etc.

Because sometimes if the 1% of a query is matching with the subject with 100% identity, then its E-value would be close to zero (very significant) but if you look at the query coverage it will be 1%. Would that be the best hit for your query? Instead, if I have another hit where 90% query is covered with 80% identity, I would go for this query as the best match.

There is an option in a BLAST program -max_target_seqs, did you try it?

[aka]

Thank you for your help, I understand I think even if I keep the best E_value and score it's not enough. No I don't tried, it allow to select the best match ?

I found this but I don't really understand:

    -max_target_seqs <Integer, >=1>
   Maximum number of aligned sequences to keep 
   (value of 5 or more is recommended)
   Default = `500'
    * Incompatible with:  num_descriptions, num_alignments

[Nitin Narwade]

I think it is very straightforward, if you select this option and provide a value of 1 (-max_target_seqs 1) it will return 1 best hit of the alignment. But it is recommended you should give a value of >5 and the default value is 500.

I would consider multiple criteria such as Percent Identity, E-value, query coverage, bits score, etc.

So if we go by this suggestion we have to process the data with the same strategy as we have discussed in the previous comment. I mean we need to sort the best hit based on multiple criteria and select it.

[aka]

Thank you, maybe it's a dumb question but how I can process my data with all this parameters? is there a way to do this sorting in the blast ?