Beginner Blasting Short Sequences

Question

[Richard]

Hi All, I have a small db of abut 100 sequencing shorter than 100bp.

I have some really short queries (5-30bp).

When I run blastn with -query myquery -subject mysubject I don't get any matches. I tried reducing the word size using the -word-size option, but still didn't get any matches even though I know they are there.

I know blast is overpowered for my little task, but I am using blast in my program for some other tasks, so it would be easiest if I could just finish it off without using another program.

Also I am using biopython for parsing the blast xml output, so if other program are suggested, please recommend something with a python parser.

thanks!

[Neilfws]

It's not so much that BLAST is "overpowered" in this case, more "inappropriate". Just because you can force it to do short query alignment doesn't mean that's a good idea. Suggest you investigate other alignment tools.

Answer 1

[Michael]

Use Ssearch36 (Smith-Waterman exact local laignments, part of FASTA tools) to get all possible matches. It is feasible with your small dataset. With the -BB option it can mimic blast text output (not XML afaik) quite well, should be good enough to use a blast parser in Bio* libraries to parse this, it worked with BioPerl at least, or you could use the FASTA-output format parser with the standard output format of ssearch. Use of any short-read/heuristic aligner is not an good option in your case, there is no reason to sacrifice accuracy for improving performance if the number of sequences is very small.

[Pablacious]

You could also use "water" from the EMBOSS package (another Smith-Waterman algorithm implementation). Totally agree on avoiding the heuristics for data sets of this size.

[Lee Katz]

This seems to be the right way to go. BLAST is in fact inappropriate due to its heuristics and the small sequence lengths. I haven't used FASTA tools but EMBOSS programs are well documented and so water should be good for beginners.

Answer 2

[Lee Katz]

I guess if they are so short then you can try to just use python to detect a 100% match in your database, ie a substring in the database equal to your query. Another option is to use a short read mapper like bowtie2, SHRiMP, or BWA and then parse the output sam or bam files.