I have some promoter motifs which range from 4 to 10 bp. I want to perform blastn search with these short motifs using 2000bp gene upstream sequences as query. the problem is that I get zero hits in local blast. I've tried using blastn-short, decreasing the word size to 4 but still i am not getting any hits. How do i proceed with this? Thank You
For short sequences, you need to also increase the E-value threshold because short sequences are more likely to occur by chance. However, I don't think blast is the right tool for sequences this short. Maybe some short read aligner would do or even regular expression depending on what you're trying to get.
I am not even sure short read aligners can handle sequences so short. They're mostly designed for the length of NGS reads which are typically longer and for alignment on large sequences e.g. genomes. If you're trying to match 4-10 bp to a 2000 bp sequence, you could just use standard pattern matching functions from your favorite scripting language.
You have to be more precise. There are many online tools with different approaches. It also depends on what the goal is e.g. discovering motifs de novo or finding known motifs and what the data is. You haven't told us exactly what kind of problem you're trying to solve. I suspect a case of the XY problem (i.e. you're asking about the solution you tried which failed instead of about the actual problem).
I am trying to find cis acting regulatory elements(CAREs) in the promoter region of some selected genes in a plant species. I took 2000 bp upstream sequence for each of these genes to search for these elements. To do this I downloaded sequences of all the CAREs ever published in plants. These CARE sequences are between 4 to 10 bases. So I want to find known motifs in promoter region of some genes. Also, if its important, genome sequence of the plant I am working on is available with very good coverage.
There are three standard ways to go about this: one I already mentioned is about finding patterns, another is to use profiles (i.e. using a position-specific weight matrix) and the last is to use HMMs.
Use of this site constitutes acceptance of our User Agreement and Privacy Policy.
version 2.3.6
How about using e.g. Bowtie2 instead of blast?
this is really nice idea, i'll download bowtie2 and try this
in fact for very short sequences botwtie (v1) is the best option rather than bowtie2 because it allows mismatches :).
I agree. Bowtie1 is a lot more sensitive than bowtie2 for such as short query sequences.
Notice that bowtie1 will give you matched sequences that are not fully identical to your subject. You will need to play with grep looking for the NM tag to control the number of truly identical annealed bases (e.g, a NM:i:0 means 100% identity, whereas NM:i:2 means that two bases are different. The concept of matching sequences and identity of these matched sequences are two different concepts in this mapping environment
I think as suggested by Jean-Karim Heriche, building HMM profiles will be a better idea than using bowtie.
Thanks for the information Buffo