Blastx similarity %

jevcampe · 05-22-2014, 08:04 AM

Dear all,

I have performed a blastx on my 30.000 contigs of my transcriptome.
I want to keep only the contigs for which the sequence description (output of blast2go) is reliable. i often see papers in which people only use a e-value cut off or a bit score cut off. Using this, I still have contigs with hits with a similarity of 40% that are not removed. I'm convinced that this is insufficient to trust the sequence description and to continue with GO, KEGG pathways and any furhter downstream application.
I would like to use a similarity cut off to reliable say that the contig is in fact the gene and related function of the blast results. Does anybody know what a good cut off value might be? Or how to proceed.
I have yet to encounter a paper where people really adress the low similarity problem. If you would have any suggestions feel free to share.

Cheers,

kurban910 · 03-18-2015, 05:21 AM

hello jevcampe,
did u solve your problem? if u did, what is the cut off you chose for similarity?

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Blastx	Desai	Bioinformatics	1	01-27-2014 11:50 PM
blastx against UniProt	papori	Bioinformatics	5	12-26-2012 07:50 PM
Learn how to use Blastx	angeloulivieri	Bioinformatics	4	07-09-2012 01:25 AM
Blastn or Blastx?	kalu	Bioinformatics	11	09-23-2011 04:18 PM
Help with BLASTX command	z3199001	Bioinformatics	3	10-11-2010 08:51 PM

05-22-2014, 08:04 AM	#1
jevcampe Member Location: Belgium Join Date: Apr 2013 Posts: 13	Blastx similarity % Dear all, I have performed a blastx on my 30.000 contigs of my transcriptome. I want to keep only the contigs for which the sequence description (output of blast2go) is reliable. i often see papers in which people only use a e-value cut off or a bit score cut off. Using this, I still have contigs with hits with a similarity of 40% that are not removed. I'm convinced that this is insufficient to trust the sequence description and to continue with GO, KEGG pathways and any furhter downstream application. I would like to use a similarity cut off to reliable say that the contig is in fact the gene and related function of the blast results. Does anybody know what a good cut off value might be? Or how to proceed. I have yet to encounter a paper where people really adress the low similarity problem. If you would have any suggestions feel free to share. Cheers,

03-18-2015, 05:21 AM	#2
kurban910 Member Location: urumqi Join Date: Jul 2014 Posts: 58	hello jevcampe, did u solve your problem? if u did, what is the cut off you chose for similarity?