Moore Notes 2 25 09

PI Conference Call

• Report approved
  • budget item breakdown needs to be modified possibly

• postdoc candidate (Morgan Langille) visiting Eisen lab Thursday

• Robin and Erick meeting
  • their pipeline is similar to AMPHORA
  • phylogenetic part is possibly better than just using PHYML
  • focus: individual sequence reads
  • application: classifying genes/gene families (gene function)
  • open problem: how to paste together into one tree?

• Steve is giving a talk in MI with one method for tree building
  • building a supermatrix and building NJ tree (faster than ML)

• Plan for Brenner meeting
  • JE idea: iterative approach (place individual reads robustly, then resolve polytomies with a different method)
  • different approach: build big joint tree (ML, Bayes, etc)
  • talk to them about computational parts, parallelization/pipeline if we knew how to do it theoretically
    • e.g. 100 gene families with 10,000 sequences each

• How/when would we move towards DNA (versus amino acid) based analyses?
  • could be helpful for
    • polytomies that are (nearly) identical at the amino acid level
    • population genetics
  • first - need full length sequences (not trimmed) rather than HMMer output - alternative method?
    • distantly related species not used to build the HMMs will not be picked up
  • then - could do codon models for tree (after back translating)
  • What data sets do we really need? Might be trying to do too much/too hard of a problem, e.g.
    • phyloCom needs tree (doesn't have to be perfect?), diversity measures might be OK too
    • might need different pipeline for making alignments for rates of evolution - need good alignments
  • Need simulations

• Novelty/potential DeRisi collaboration
  • evolve new trees and evolve new sequences on these trees
  • Olivier Lichtarge
  • Tim Cooper
  • There's a big DARPA project in this area

• Need to reassess future directions