Hello,
I have nearly 300M barcode sequences, length of which is about 16 nt. For each of them I want to find out the best similar target from the whitelist database containing about 700 thousand sequences with comparable lengths.
To do that, I Initially tried bowtie or bwa, which failed to work due to the short lengths of references. Then, I tried blastn by spliting barcode sequences and run batch blast tasks parallelly. However, it's still time consuming.
Does anyone have better ideas to accelerate that process?
Thank you very much!
Hello,
You can check MMseqs2: https://github.com/soedinglab/MMseqs2
It is design to work with hudge datasets and it should be much faster than Blast. My only concern is that 16 nt is a very short fragment and therefore you might have problem with sensitivity, and for that short fragments you will likely need to adjust the e-value threshold
Dear Jakub,
Thank you for your kind advice. I will track MMseqs2 ang try it then. To my view, sensitivity should be partially improved if word size and penalties for mismatches and gaps could be modified in MMseqs2.
- Badges
- Science topic