blastn no hit found for very short sequences

Question

[exp.sherry]

I am trying to blast a set of primers (I picked from reference genomic sequences), but for all of the short primers (<= 15nts), I got "no hit found" with the command

blastn -db db/GRCh38p7toplevel.fa -query sample.fa -out results.txt -task blastn

(GRCh38p7toplevel.fa is build with all primary assembly of human chromosome sequences) but on the web version, I got lots of hits. Also, if I remove the "-task blastn" by the end, a lot of primers gives "no hit found"

[GenoMax]

Since you are using short queries you should try -task blastn-short with your search. 15 nt is getting awfully close to default word size in blast for nucleotides.

If you are just checking your primers you could also try in silico PCR primer tool at UCSC.

[exp.sherry]

Thank you! it works! I am doing a multiplexing pcr(about 100 amplicons) in one tube, it becomes almost impossible to check all combinations of primers use the pcr primer tool, so I switched to blast. Is there any other way to check this ?

[Denise CS]

The option provided by @genomax2 seems to have been reported to work for sequences > 17 nt (inclusive) according to a previous Biostars post. This other post and the NCBI Options for the command-line applications may be worth checking too.

Answer 1

[Jean-Karim Heriche]

Blast is a heuristics. Its results depend on the choice of parameters. For short sequences, you need to reduce the word size to about half the query length but also increase the E-value to see more hits with low p-value because short sequences are more likely to occur by chance. This paper has a table showing Blast misses as a function of word, query size and mismatches.