Moore Notes 1 9 13

Group Call

• Participants: Katie, Jonathan, Stephen, Tom, Guillaume, Josh, Dongying

• Progress report

• Hiring
  • Sarah Hird postdoc candidate for Eisen lab
  • UCSF bioinformatics hire - final candidate interview Friday

• Data opportunity from Declan Schroeder
• 18S/16S, but not shotgun metagenomics

• Tom: SFam updating/fixes
  • Automating and improving the process for family construction
    • LAST vs. BLAST
    • Sift with LAST first and then use HMMs
    • Stephen's additional ideas
  • Do we need supercomputer time?
    • Cluster access probably OK for now (2x per year updates)
  • Incorporating metagenomic sequences in addition to genomes
    • Much more complex and larger data sets (so might need more compute resources)
    • Youssef has published methods
  • Future of SFams
    • Some one else (CAMERA, JGI) might be able to do the searches that we need
    • We care more about having good families more than maintaining a protein db
    • People from Banfield and Giovannini labs are using SFam HMMs
  • Targeting next SFams release in a month or so

• Josh: beta diversity mapping
  • Method
    • Model community dissimilarity (difference in OTU abundance) as a function of distance in environmental variable space
    • Make predictions using raster data
    • Project predictions into geographic space
    • Average in 100km window around each grid cell
  • Preliminary results
    • Highly correlated with primary productivity
    • Tom: relative abundance metric might be affected by sampling issues
    • Katie: what about seasonality? Josh: All subsets regression selected annual variables over month-specific ones, but will try putting month in the model to see what month-specific predictions look like
    • Cross-validation is pretty good
    • Declan's data might be good for validation
  • Community classification
    • Clustered a random subset of grid cells, applied kNN algorithm to assign remaining cells to clusters
    • Arbitrarily picked 6 clusters
    • Methods could be improved
    • Result: Arctic different from Antarctic
  • Working on plotting direction of greatest turnover at each grid point
    • JE: Sounds similar to source-tracker software (Rob Knight et al.) http://www.nature.com/nmeth/journal/v8/n9/abs/nmeth.1650.html
  • What are good applications for this?

• New paper form Noah's group http://www.pnas.org/content/109/52/21390.full