Moore Notes 5 13 09

Group Call

• JE just got a budget from Gail
  • PIs will follow up on this later
  • Also need to talk about hiring

• ASM
  • JE and JG will be there
  • Kelly will too
  • OK to talk about our methods in development?

• Building trees from large alignments
  • Steve: concatenated alignment of 31 markers (~7000 aa long, ~5000 reads, 570 genomes-AMPHORA)
    • Trying to build tree
    • FastTree works
    • Wants to use likelihood
    • Memory issue? Test with a debug tool
    • Trying using FastTree (or genome tree) as a constraint in RAXML
  • Martin: used phyML for whole genome tree
  • Try bowtie? only for human...
  • JE: ask Sanderson's group
  • Any reason not to use a guide tree?

• Simulator is revised
  • Need to grow data sets to get bigger
  • Subsamples from (a subset of) AMPHORA db
  • Need input from Srijak

• Binning/OTUs: Kunin paper
  • Extrinsic approaches
    • ref genome closely related to reads, BLAST, pull out those reads (as in Rusch paper)
  • Intrinsic approaches
    • assemble (a little)
    • nucleotide composition: works for low complexity ecosystems
  • Which to use depends on what you want to do, e.g.
    • how does number of bins grow with spatial scale? Taxonomic classification -> AMPHORA on markers works here
    • to estimate total number of taxa in all oceans ->will AMPHORA underestimate? better with more markers
    • are pairs of reads from the same taxa?
    • move to DNA sequences to resolve fine level distinctions
    • no individual gene will get the tail of the distribution (rare organisms)
  • OPFs/environmental gene tags
  • alternative: 454 sequence 1 million rRNAs from 1 sample (deep coverage) vs. metagenomics
  • Plan for GOS
    • markers plus more genes
    • use reference genomes
    • iterative algorithm based on a guessed set of dominant taxanomic groups
  • follow up skype call: Josh, James, Dongying, Jonathan