Using alignment tools for unintended tasks.

[Vilhelm84]

Hey all. I have a project in which we are sequencing off a series of transfected plasmids each with a series of mutations. In sample prep we made used plasmid specific/illumina primers to make our sequencing libraries, one off the input plasmids, one off the resulting RNA transcribed from the transfected plasmids. We sequenced both libraries using the Hi-seq platform. Basically the idea is to map millions of 100nt reads that have 0-9 mutations to the same short reference sequence (the original un-mutated plasmid).

I am using bowtie to map these reads. Since these reads are only needing to be mapped to a few exons found in the plasmid, I used the "wild type" non-mutated plasmid as my reference "genome". When I align these reads however, any reads that came from mutants with 3 or more mutations will not be mapped due to the settings found in the software.

I realize that all the deep seq tools have been designed to 1. map short reads over the whole genome quickly and 2. filter for reads that have multiple mutations since these are generally bad reads. But in my case, my reference sequence is tiny AND I want to be able to see the reads that have multiple mutations. Is there a way to align reads with multiple mutations using bowtie or tophat? Is there a program that works better for this sort of alignment (remember 54 million reads)? Preferably a program with similar amounts of documentation to bowtie or tophat as I am a newbie bioinformatician?

Thanks, Will

[Richard Finney]

Try Jim Kent's blat program from UCSC.

http://genome.ucsc.edu/FAQ/FAQblat.html

[twaddlac]

I like using nucmer (http://mummer.sourceforge.net/). You have to do a lot of manual parsing depending on your analysis, but I generally use a database like psql in order to figure things out about my data. It provides percent identity which you could use to deduce an approximation of the number of mismatches. Alternatively, you could use the show-snp and show-tiling programs that come along with it.

Also, I think bowtie2 is more tolerant to mismatches (though they do restrict the number of mismatches in the seed). So if your mismatches are spread out enough I think this might work.

You could also try SHRiMP (http://compbio.cs.toronto.edu/shrimp/) which isn't limited by number of mismatches and provides the ubiquitous SAM/BAM output format.

There are many other mappers, but these are some of the easier ones that I've encountered and they all seem to be pretty accurate.

If you're interested and up for a challenge, check out bfast and bwa. They're tough to run but the authors seem to be very helpful.

I hope this helps!

[mchaisso]

You can use either bwa-sw or blasr (google blasr github). bwa-sw can tolerate many errors, and blasr is used to map illumina reads to pac bio reads to error correct them, which may be similar to the case that you have.

-mark

Hey all. I have a project in which we are sequencing off a series of transfected plasmids each with a series of mutations. In sample prep we made used plasmid specific/illumina primers to make our sequencing libraries, one off the input plasmids, one off the resulting RNA transcribed from the transfected plasmids. We sequenced both libraries using the Hi-seq platform. Basically the idea is to map millions of 100nt reads that have 0-9 mutations to the same short reference sequence (the original un-mutated plasmid).

I am using bowtie to map these reads. Since these reads are only needing to be mapped to a few exons found in the plasmid, I used the "wild type" non-mutated plasmid as my reference "genome". When I align these reads however, any reads that came from mutants with 3 or more mutations will not be mapped due to the settings found in the software.

I realize that all the deep seq tools have been designed to 1. map short reads over the whole genome quickly and 2. filter for reads that have multiple mutations since these are generally bad reads. But in my case, my reference sequence is tiny AND I want to be able to see the reads that have multiple mutations. Is there a way to align reads with multiple mutations using bowtie or tophat? Is there a program that works better for this sort of alignment (remember 54 million reads)? Preferably a program with similar amounts of documentation to bowtie or tophat as I am a newbie bioinformatician?

Thanks, Will

[lh3]

If you have a very short reference genome, I would recommend ssearch - a pure smith-waterman algorithm, though powered by SSE2.

[rskr]

bwa you can set the maximum mismatch to 9-10 it will be really slow since the search space grows exponentially per-mismatch, also the accuracy might suffer. Another alternative, is that if you have an idea of what the mutations are, you can create a database with the variations of the mutation in it.