[mira_talk] Re: Algorithm of MIRA to assemble 454 reads

On Friday 10 October 2008 06:28, Toshiaki Namiki wrote:
> Does anyone know about the algorithm of MIRA to assemble 454 reads?

Hello Namikisan,

yep, I do :-)

> I'd like to study about the assembly algorithm of MIRA, and I found the
> thesis. ( http://www.chevreux.org/uploads/media/chevreux_thesis_MIRA.pdf )
> But it was written in 2005, so I don't think it refers to 454 reads.
> Can the thesis be applied to 454 reads?

Yes. And no. The basic concepts are still valid (iterative approach, different 
filter and overlap stages, pathfinder & contig interaction, automatic editor, 
knowledge buildup of repetitive regions etc.pp). A few things have been 
added - like a robust overlap graph pruning - but these are not changing the 
overal strategy.

The "big" change is that most of the modules are now aware of rather very 
different sequencing technologies (Sanger vs. 454 vs. Solexa etc.) and that 
they need to take sometimes rather different strategies to solve a problem. 
Just as example: the problem 454 has with homopolymers led to all kinds of 
changes like, e.g. routines that search and tag misassembled repeats are not 
allowed to take decisions for columns involving gap bases. There's a new 
automatic editor for 454 data and other routines were added that take into 
account whether homopolymer sites are in the vicinity of problematic bases to 
take decisions.

Another example are the problems caused by Solexa data: starting with the 
sheer amount (between 3.5 and 8 million reads just for one lane) and 
continuing with the technology specific error distribution, new algorithms 
that cope with those specific problems were added to the modules.

There will be an updated description for the assembler available, but I will 
need time to write it and this is in conflict with the need to improve the 
assembler itself so that it can cope with the kind of projects I am working 
with right now.

I'd need 12 additional hours per day :-)

Regards,
  Bastien

-- 
You have received this mail because you are subscribed to the mira_talk mailing 
list. For information on how to subscribe or unsubscribe, please visit 
http://www.chevreux.org/mira_mailinglists.html

Other related posts: