Compare fasta files

[Hel]

Hi there,

I want to compare several fasta files containing sequences. These sequences are transcripts obtained from RNA-Seq. I want to find out the shared transcripts between samples.
I cannot use CuffCompare or similar because I have no reference genome. I only have transcripts.

Thanks in advance,

[dschika]

You could try CD-HIT to cluster the reads.

[Richard Finney]

Try using command line utilities
cat
sort
uniq


example :
#get unique reads for 1, filter out read names (lines with >)
cat 1.fa | grep -v ">" | sort | uniq > 1.tmp
#get unique reads for 2
cat 2.fa | grep -v ">" | sort | uniq > 2.tmp
#get reads common to 1 and 2
cat 1.tmp 2.tmp | sort | uniq -d


sort takes a "more RAM memory" parameter if it's a large data files.
check out the manual using "man sort" for details

[vivek_]

BLAT appears the easiest and straightforward way right?

[Richard Finney]

Check out bl2seq ...

http://blast.ncbi.nlm.nih.gov/Blast.cgi?PAGE_TYPE=BlastSearch&BLAST_SPEC=blast2seq&LINK_LOC=align2seq

There's a command line version if your into that kind of stuff.

[Hel]

Thanks to all, I am very grateful for your help,

This is my opinion:

(i) CD-HIT seems interesting, but I have not test it yet.

(ii) Using 'cat', 'sort' and 'uniq' commands in FASTA files performs an enormous mistake, because it compares each line instead of each sequence (composed of multiple lines).

(iii) 'BLAT' needs a reference genome, and I do not have such.

(iv) 'bl2seq' does not support large size files, so that they suggested to use Blast+. So it is the same as running a local blast with Blast+.

Is this correct?

[GenoMax]

(iii) 'BLAT' needs a reference genome, and I do not have such.

Blat does not need a reference genome. In fact you use blat with just two files (which can be single sequences or multi-fasta files). First file on the command line serves as the "database" and the second "query". So in your case you will be blatting a sequence (actually many of them sequentially) against one "database" file (or the whole lot of files concatenated together). Ideally the sequence itself will be the top hit. You may want to use tabular format to be able to parse the results easily.

[rhinoceros]

(ii) Using 'cat', 'sort' and 'uniq' commands in FASTA files performs an enormous mistake, because it compares each line instead of each sequence (composed of multiple lines).

You could remove the linebreaks in seqs and then continue as Richard advised..

awk '!/^>/ { printf "%s", $0; n = "\n" } /^>/ { print n $0; n = "" } END { printf "%s", n }' file.fa > out.fa