building a Mosaik reference for Mouse

[sdriscoll]

Can anybody point me in the right direction for building a single, full genome, reference for the mouse that i can then align my illumina read data to? I have FASTA reference files, one per chromosome, for the mouse which I downloaded from UCSC. I can pass one of those at a time to MosaikBuild to produce .dat files for each chromosome but that seems a little crazy because that means I'd have to run a single lane of data against each chromosome, 1 at a time.

If this is how other people do it then that's totally fine - it just seems like I should be able to build a single reference file for the entire genome.

[snownebula]

Hi there,

All you have to do is create a concatenated FASTA file and you'll be all set with MOSAIK.

For example, if I wanted to combine the first four mouse chromosomes into one file, I could type:

cat mm_ref_chr1.fa >> mouse_ref.fa
cat mm_ref_chr2.fa >> mouse_ref.fa
cat mm_ref_chr3.fa >> mouse_ref.fa
cat mm_ref_chr4.fa >> mouse_ref.fa

You could keep doing this for all of the mouse chromosomes or if you're savvy at creating bash scripts, you could pretty much automate the above in a small script.

Cheers,

// Michael

[sdriscoll]

cool thanks. it just wasn't clear in the documentation that you could just cat files together to make one larger reference. now i just need to figure out this jump database thing and i'll be off and running. Mosaik chews up some serious RAM and i've only got 16 GB on the system I'm running it on. looks like a jump database will help for running full genome alignments on this system.

[sdriscoll]

so i made this cat'd reference file (2.6 GB) and compiled it down. then i made a jump database and started a run with MosaikAligner using the jump database of this full genome reference. i ran pretty much the default settings listed in the manual except with only 4 cpu cores. it looks like it munched up about 19GB of RAM to load the jump database files into memory but once the alignment actually started it wasn't using all 4 cores - it was only using about 4% of the CPU and it was processing only 3.5 reads per second with an ETA of 53 DAYS. what could be going wrong?

[donniemarco]

cat all files

maybe concatting different files might get little tedious. i tried:
cat chr*.fa >> human_ref.fa

it worked well.

[mkeehan]

Are you still using your system with 16GB of RAM?
You are probably swapping if it's using 19GB...

I found reading the manual to get the right parameters made a huge difference to the reads per second. The magic parameters I found were
-bw 13 -act 20 -mm 4 -mhp 100
That took me from a few reads per second to 700 - 800 per second.

I also needed around 20GB of RAM for the jump database.

[sdriscoll]

thanks for sharing the magic.