Moore Notes 8 5 09

Group Conference Call

• Martin: still waiting for subcontract
  • Tom: JE worked on it Monday

• Josh: update on GOS
  • Met with James in Eugene
  • Emailed Rusch for pooled GOS reads assigned to OTUs, but no response
  • James: what file were they working from?
  • Plan:
    • run all reads from all samples through MOTHUR
    • map OTUs back to samples
  • Has any one heard from Jenna? Busy with course.

• Tom: CAMERA db access update
  • Every one has a login
  • One person per lab has a db account
    • idea is to share account information
    • account person is the point person to communicate with CAMERA
    • db schema is changing, but will be more fixed soon

• Srijak: new file server set up at Davis
  • Will send out info by email

• Tom: svn repository
  • Tried to contact Mike Lewis re: repository on genbeo
  • Talked with JE/Srijak: will do it on the new file server
  • Srijak and Russell will work on it
  • svn (centralized) vs. git (redundant)?
  • For what: sharing
    • data?
    • software

• Dongying question: HMM profile search against genbank
  • how to get a cut-off for bit score? normally distributed?
  • Tom: doesn't scale well across all types of sequences
  • How is it related to the E-value, extreme-value distribution?

• Josh topic: randomized (or gridded) sampling
  • locations of samples in GOS and other data sets are not random
  • difficult to make inferences about distance decay and other ecological inferences
  • may also affect estimates of global phylogenies (constructing them, relating them to geography)
  • Can we identify a data set where sampling was done randomly with respect to geographic location (or time)?
    • if not random, a grid might be OK
    • Tom: mycological data from a grid layout in cane fields - in the works (~1 year) from UC Berkeley
    • Katie: Jed Fuhrman's data?
    • Martin: International Census of Marine Microbes (ICOMM, Mitch Sogan), 16S only?
  • James: Could we estimate the error in using non-random data?
  • Josh: We could write a paper arguing for designing studies with random sampling
    • Check with simulations if it is a big problem or not, and how bad, and for techniques?
    • Can you fix it? Or estimate your bias? This would depend on assumptions about spatial distribution.

• Tom: metric for measuring taxonomic & genetic diversity between communities
  • looking for gene family with phylogeny that is different between communities
  • approach: compare phylogenetic diversity measure
    • but, this can be confounded by differences in taxanomic diversity
    • wants a metric to "subtract out" taxanomic diversity from gene family diversity
  • James: Helene's paper relating total phylogenetic diversity (branch length of tree) vs. taxanomic diversity
    • power law
    • working with 16S
    • may want to try other gene families
  • Josh: another approach is to write down a statistical model with variance components for different contributions