How to decide whether a hit is unique or not?

[Jamawoko]

Hi, maybe it is a trivial question but I did not find a satisfactory answer at the forum. How can you decide by score or other evaluation values whether a hit is unique or not?

I have this example (from BLAT output):

matches

query size

chromosome

2440

2524

11

31

2524

10

Is that a unique hit because the matches of the first hit are much more than the matches of the second hit?
For example, is there a rule that you can say the hit is unique because the score of the second best hit is lower than 90 % of the best hit score or something?

[GenoMax]

What BLAT parameters were you using for this search?

[Jamawoko]

I used Ilumina data, did a bit cleaning with fastq, assembled it with Inchworm RNA-Seq Assembler, deleted all contigs which were short than 200 and did a run with gfClient from BLAT package. The input was a fasta File and I aligned it against a nib directory containing all human chromosomes. I used no further parameters.

[Richard Finney]

Select a cutoff for matches; i.e. matches must be a certain percentage of query size, e.g. 96%. Additional filters would be number of inserts and count of inserted bases (if genomic, don't use if RNAseq and splices expected). Playing around with the cutoffs should give you the answers you want.
This is easy to do by writing a simple awk or perl script. The exact cutoffs you want are going to be based on the noise level of your data.

After filtering you can determine whether your read maps to only one location on the genome by counting the number of occurrences of the read in your psl (blat output format) output.

[GenoMax]

I used Ilumina data, did a bit cleaning with fastq, assembled it with Inchworm RNA-Seq Assembler, deleted all contigs which were short than 200 and did a run with gfClient from BLAT package. The input was a fasta File and I aligned it against a nib directory containing all human chromosomes. I used no further parameters.

If you used the gfClient then these are the parameters you likely used: http://genome.ucsc.edu/FAQ/FAQblat.html#blat5

Default settings for the blat program for other options are described on this page: http://genome.ucsc.edu/goldenPath/help/blatSpec.html

Richard Finney has already given some pointers on what parameters to change.

[Jamawoko]

Thanks to both of you. Your answers helped me a lot.
So you mean there is no specific general rule to decide whether a hit is unique or not. It all depends on how you set the program parameters you want to make referring to the data you are analysing.