I'm in a situation where I have two sets of BLAST results, and I'd like to compare them (non-rigorously). What is the best way to average the evalues (or bit score, whichever is easier) of the HSPs in a BLAST report? I'm basically looking for a way to compare length-weighted results.
•
written
by
David M
1 answer
I agree with Eric and Istvan - both relevant points.
I'd start with a standard parser to put the BLAST results in a table. This will allow you to make many types of comparisons. Without knowing the details of what you're looking for, it's hard to give detailed advice. A distribution of values or scores (where those values have been binned, eg E-value bins to 0.001, to 1E-06, 1E-10, 1E-20, 1E-40, 1E-100 (to be somewhat arbitrary)) might be more interesting. You can also look at something like how many BLAST HSPs (essentially the alignments returned in the results) contain residue 1 to residue n of the query - something like "coverage." So, there you have two rather different ideas to take up, again, depending on what you're looking for.
Log in to answer this question.
-
Comparing DEG lists between clusters, looking for unique/differences in pathways between them
written by Michael •
-
Best hit from BLAST for a particular genomic region
written by johnny.sf •
-
Advice on Blast output
written by stacy734 •
-
If there are two or more B genes with same highest bit score in RBH method
written by madapada •
-
Understanding Best-Hits filtering algorithm in BLAST+ applications
written by Denis
-
Comparing Quality Of Read 1 And Read2 Of Illumina Pe
written by Juliofdiaz
-
Statistics: Getting The Standard Error For Expression Level Fold Change, Based On Geometric Averages
written by David M
-
Ncbi Blast Output Missing Subject Name
written by Razzmataz •
-
Comparing Protein Datasets By Sequence
written by Kevin
-
Best Blast Hit Without Pain
written by kpatil •
Averaging evalues or pvalues does not make a whole lot of sense, whatever the application you have for it. Maybe the percent of similarity (Identical bases / total number bases) would be more interesting? I think describing what you are trying to accomplish (on a broader scale) may sprout more useful suggestion. Cheers
You will need to define what you mean by 'comparing' the BLAST results. What are you looking for?
Sorry for the lack of detail. I've got two BLAT hits (full gene found on 2 different contigs) for one putative gene (each BLAT hit is made up of ~8-14 'exons'). I'm wondering which of these hits is closer to the gene I'm searching for.
Wait, have you got BLAT or BLAST? With BLAST, a direct way of comparison is the bitscore, in BLAT you may want to refer on the calculated coverage and identities, so you can distinguish between global and local similarities.