JasonStajichHistory < Blog

Blog > JasonStajichHistory

TWiki webs: Main | TWiki | Sandbox Log In or Register

JasonStajich on June 24 2005 @ 12:31pm
- MartinSenger - BioNanny
  - WebServices monitoring.
  - measure how long webservices took. Metadata returned.
JasonStajich on June 24 2005 @ 12:08pm BOSC Blog, Lightning Talk
- AlanChristoffels - Bio.Mambo
  - Bioinformatics portal. http://www.tll.org.sg http://www.tll.org.sg/res/alan_pub.asp
  - Parallel data distribution
  - Cluster management
- VectorNTI integration with BioPerl
  - Development environment within windows environment
  - VectorNTI API, allow use of the VectorNTI framework, COM based...
  - VectorNTI is commited to making tool available to "Non-Profit" Researchers, Free software licenses, paid tech support.
JasonStajich on June 24 2005 @ 12:06pm
- BioNote a Wiki for molbio lab folks
JasonStajich on June 24 2005 @ 7:26am
- AaronDarling's mpiBLAST talk
- Parallel BLAST, parallel allows it to scale the aggregate cluster memory
- Superlinear speedup in terms of performance
- 0.9 version
  - Accuracy was limited
  - poor fault tolerance
- 1.0
  - Accuracy is still limited
  - better useability and output formats
  - Scalability is actually reduced!
- 1.2.1
  - "distributed biosequence lookup"
  - packages
  - Improved scalability
  - Bad accuracy (still)...
- 1.3
  - Pipelining results
  - Exact e-value statistics
  - Pipeline database fragment copy (instead of all copying from NFS at same time)
  - Scalability is actually worst (so 128 nodes search is worse than single)
- mpiBLAST-g2 (ported to GLOBUS architecture)
- pioBLAST - my NCSU/ORNL -- single database image, run on a parallel filesys,
  - Uses parallel IO, pre-partitioned database
  - Scalability graph linear
- mpiBLAST 1.4.0
  - Fix distributed suquence lookup
  - Fix framgment copy algorithm
  - So basically should be more scalable
  - 305X speedup on 128 Nodes
JasonStajich on June 24 2005 @ 6:51am BOSCBlog
- http://www.affymetrix.com/support/developer/tools/download_igb.affx
- [GregHelt]'s talk at BOSC, about DAS/2
- New spec to deal with more explicit models
- DAS/2 complete within a year
- Runs off of the GenericGenomeBrowser
JasonStajich on 11 Jun 2005
- New page for looking at pairwise orthologs
- To Do: Improve pairwise alignment viewer to show gene structure snapshot - as FTstring text and then as GenericGenomeBrowser image - need to make it a very small image though.
- Write homolog browser to display summary results and links

Testing molecular clock in PAML.

Basically this code for doing an en-masse comparison. This is as per MBL MolEvo Course slides

# $type would be either 'clock' or 'noclock' depending on which file you are reading.
# so you'd want this type of loop to run over all the result files from PAML
# in my case I've numbered all the gene clusters and that number is stored in $num
# designed to parse PAML mlc files from codeml
while(<IN>) {

  if( /lnL\([^\)]+\):\s+(\-?\d+\.\d+)/ ) {
     $dat[$num]->{$type} = $1 * -1; # this is just a lnL it is not -lnL
  } elsif( /^ns\s+=\s+(\d+)/ ) {
     $dat[$num]->{'ns'} = $1;
  }
}

# to then compare the clock vs no-clock
my $i = 0;
print join("\t", qw(CLUSTER NOCLOCK CLOCK LRT SIGNIF)),"\n";
for my $d ( @dat ) {    
    if( defined $d ) {
        # H1,H0 = - log (likelihood)
         
        # LRT = 2*(lnL1 - lnL2);
        my $lrt = 2 * ($d->{'clock'} - $d->{'noclock'} );
        if( ! defined $d->{clock} || ! defined $d->{'noclock'} ) {
            warn("no dat for $i\n");
        }
        my $p = Statistics::Distributions::chisqrprob ($d->{'ns'} - 2,$lrt);
        printf "%s\t%9.3f\t%9.3f\t%-7.2f\t%-5.2g%s\n",$i,$d->{'noclock'},
        $d->{'clock'}, $lrt, $p, $p < $signif_high ? '**' : $p < $signif_mod ? '*' : '';
    }
    $i++;
}

-- JasonStajich - 08 Jun 2005

[GMOD] Progress. The links to the presentations from their May meeting are here

There is a new Architecture discussion group to hash out the ways to make the tools interoperate better. Hopefully these efforts will establish some standards and leadership for project coordination.

-- JasonStajich - 08 Jun 2005

Almost finished setting up my new data site for fungal genomes. http://fungal.genome.duke.edu

This site has a BLAST interface which uses my Bio::SearchIO parser and Writer (BioPerl) to re-write the report with links in to GenericGenomeBrowser.

There is also gene retrieval CGI direct from Bio::DB::GFF which I think is pretty cool.

-- JasonStajich - 03 Jun 2005