Can BLAST do the RNA-seq reads mapping efficiently?

[nivea]

Hi all,

I'm dealing with low-quality reads of a bacteria which are 100 base long. Thus, most of the software cannot map back most of the reads to the reference genome with no more than three mismatches. So can I use BLAST to do the mapping work? Will it to slow?

Thanks

[Thorondor]

amount of reads? size of your reference genome? what software you tried already?

[Simon Anders]

if you think you have more than three mismatches, why do you limit your aligner to it? (Your questions sounds a bit as if you think that most short-read aligners had a hard-coded limit to the maximum number of mismatches allowed. They don't.)

[Thorondor]

if you have illulmina reads you might want to read this review:
http://seqanswers.com/forums/showthread.php?t=11045

[doc.ramses]

Originally Posted by Thorondor

if you have illulmina reads you might want to read this review:
http://seqanswers.com/forums/showthread.php?t=11045

Can someone send me the paper via PM ?
Thank you !

[nivea]

Originally Posted by Thorondor

amount of reads? size of your reference genome? what software you tried already?

Hi Thorondor,

I used bowtie to map32129789 reads (76bp) to the bacteria E.coli K12 with a tolerance of mismatches <=3. However, only 0.02% of the reads mapped uniquely to the genome.
After using BLAST check the result I found for most of the reads, only the previous 35b of reads got hit in the genome. Thus, it's probably because of the sequencing error that I cannot use bowtie to map them back.

So I want to try BLAST. But do you think it's too slow? Or does this program work accurately enough for reads mapping?

Thanks for reply~

[nivea]

Originally Posted by Simon Anders

if you think you have more than three mismatches, why do you limit your aligner to it? (Your questions sounds a bit as if you think that most short-read aligners had a hard-coded limit to the maximum number of mismatches allowed. They don't.)

Hi Simon,


The problem is the quality of the reads, which you can see from the upper post. The sequencing error made only smaller than 35bp of most of the reads can map back to the reference genome. If I want to take advantage of the current reads, do you think it's good enough for me to use BLAST?

Thanks~

[Simon Anders]

Actually, Bowtie should disregard mismatches on low-quality bases, but this does not work that well.

However, why don't you simply trim off the bad-quality parts and give the trimmed reads to Bowtie? In a bacterial genome, 35 bp should be more than enough to get good matches.

regarding BLAST: It is way too slow for millions of reads, and its model of scoring mismatches is based on assuming mutations rather than read errors to be the cause, which is inappropriate.

[Thorondor]

so why not trim the reads first and then try to map without quality values? You don't want to assemble bad reads anyway and they mostly won't overlap if you don't trim. Or you could just cut all your fastq reads to a length of 40 bp and then try to map with BWA.

Maybe you also want to have a look at Cufflinks for your assembly? http://cufflinks.cbcb.umd.edu/

edit: simon was faster :-/

[nivea]

Originally Posted by Simon Anders

Actually, Bowtie should disregard mismatches on low-quality bases, but this does not work that well.

However, why don't you simply trim off the bad-quality parts and give the trimmed reads to Bowtie? In a bacterial genome, 35 bp should be more than enough to get good matches.

regarding BLAST: It is way too slow for millions of reads, and its model of scoring mismatches is based on assuming mutations rather than read errors to be the cause, which is inappropriate.

Thanks Simon! That's very helpful.

[nivea]

Originally Posted by Thorondor

so why not trim the reads first and then try to map without quality values? You don't want to assemble bad reads anyway and they mostly won't overlap if you don't trim. Or you could just cut all your fastq reads to a length of 40 bp and then try to map with BWA.

Maybe you also want to have a look at Cufflinks for your assembly? http://cufflinks.cbcb.umd.edu/

edit: simon was faster :-/

Thanks Thorondor!