Dear Bioinformaticians,
I would like to ask about defining the level of filtering by sequence identity (pident, %) from tblastn results.
I have a table of tblasn results in Galaxy including about 800,000 sequences. I would like to filter them by sequence identity but if I filter them with 98% I lose almost all sequences. I would like to know what is the accepted level for filtering considering that this is from protein! data. I think this should not be as strict as a blastn filtering (commonly 98 or 99%). Please give me advice and link me to any publication which tells me a proper percentage.
All answers are greatly appreciated. :)
Thend
1 answer
Hi priyankamowlali,
We used 96% minimum percent identity, later we defined another threshold of 98%. Although we lost majority of our sequences we made sure that they are good/high quality. I hope this helps.
Best, Thend
Log in to answer this question.
-
BLAST search with multile fasta sequences and with a single output
written by endretoth •
-
How to filter BLAST results by query coverage? How to analyze large BLAST outputs?
written by braun_tube •
-
Filter fasta sequences based on a list with grep or BBmap
written by endretoth •
-
Filtering tblastn results
written by juan.crescente •
-
What is the difference between blastx and tblastn?
written by Kumar
-
Similarity in particular gene/sequence of different phylum
written by nimiumretiya12 •
-
Batch extracting various coordinates of various FASTA IDs directly from NCBI?
written by yotiao •
-
Help with CD-hit commands for DNA sequences
written by frcamacho
-
Help to filter blastp results
written by guillaume.rbt
-
Rows Comparison In Excel Using Perl,, Sequence
written by Kiran
It is impossible to say without knowing any details about the project. Why do you need to filter the sequences?
Even when knowing the details, there is probably no perfect threshold, it is often a trade-off between removing artifacts (I guess this is what you want to do) and not losing too much information.
hey endretoth, I am doing a similar work Can you tell me how did you filter the sequences ? manually or did you use any programming language to do that?
It depends on what kind of sequences you are using in search. you should also be looking for the blosum or PAM matrix you used in blast, depending on divergence between the sequences you are looking for and query sequence.