Does this fit to your interest ?

https://www.biostars.org/p/241972/

Abhijeet Singh thank you for answer, yes it fit my interest, but it does not offer any user-friendly way of solving the problem

Guitope

http://www.bioinformatics.org/sms2/shuffle_protein.html

Sagar Bashyal, thank you for answer, shuffle protein is useful tool, but I want to downsample set of sequences, not to shuffle order of their amino acids.

1. Generate a lists of 5-digit random numbers using an online random number generator, such as,

https://www.random.org/sequences/

12010

15534

17276

......

2. Add a letter P in front of each number (this can be done in Notepad++ by replace each line break (\n) with (\nP).

P12010

P15534

P17276

......

3. Treat the results as Protein IDs, and pull them from UniprotKB (or any other protein databases)

http://www.uniprot.org/uploadlists/

you will get a list of random protein sequences:

>sp|P12010|LEU3_BACCO 3-isopropylmalate dehydrogenase OS=Bacillus coagulans OX=1398 GN=leuB PE=1 SV=1 MKMKLAVLPGDGIGPEVMDAAIRVLKTVLDNDGHEAVFENALIGGAAIDEAGTPLPEETL...

>sp|P15534|SCR3_NOTGO S-crystallin 3 (Fragment) OS=Nototodarus gouldi OX=6631 PE=1 SV=1 PAPNYTLYYFNGRGRAEILRMLFAAAGQKFNDKXXEFN...

>sp|P17276|PH4H_DROME Protein henna OS=Drosophila melanogaster OX=7227 GN=Hn PE=2 SV=3 MYQRQVSFDKPTRVEDSAYIVEGVDIKEARNTCLLFSPKDSSLSSGALANILAIFKKHDI...

Does this meet your needs?

Thank you for great answer, Xiaolong Wang,

it looks like ingenious solution :-)

@Wang - Can you justify your answer ?

Is this random number generated by you, is purely synthetic ? Because here its not what he wanted to do. He wanted to randomly choose the sequences from his sequences. By generating pure random number, its like you are gambling, if the random number matches the protein accession in uniprot, then you can get the hit otherwise not. I guess the question was to randomly select the sequences present in his fasta files.

@Martin

In this case, I would use "grep" (http://grep.3tecky.cz/) to extract the random sequences with any defined regular expression from fasta file. And then pull those sequences from any data base, here Uniprot.

Thank you, Abhijeet Singh,

the previous answer really does not solve my problem, because I am interested in random sampling from set of defined sequences (for example sampling 100 sequences from 20 000 human protein sequences). I don´t realize this before. Your solution is good, thanks.

@Abhijeet Singh

Yes, your method is great! Cool!

@ Martin Bartas

My method do have a problem for a user-defined database. Although it can surely be modified to meet the requirement, but it is uneccessary now, the above method is good enough.

I do not know there is a online grep tool. It is a little bit strange to me, in what situation people need an online grep tool? Why not using command line grep instead, even for teaching students？

Thanks!