Moore Notes 7 6 15

Group Call

• Participants: Katie, Josh

• Progress report submitted

• ShotMAP paper submitted

• Josh: Tara Oceans - preliminary analyses
  • Spatial autocorrelation measured by Morans I statistic
    • About 3% show autocorrelation
    • Example of a KEGG Orthology Group that shows autocorrelation (map)
    • Distribution of I values in broad functional categories: plot
    • Distribution of statistical significance with Bonferroni correction: p-values using MCMC simulations
      • Patrick: Try FDR correction
  • Methods for quantifying KO abundance
    • Fragment recruitment to assembly based database (Tara Oceans) versus classification to a protein database
      • Patrick: How about first assembly, then genomic protein db? Stephen: humann2 does this with genome db then genomic protein db
    • Tom: compare to RPKG values with AGS normalization
    • Patrick/Stephen: What about abundance? Or absolute abundance (i.e., concentration)? Josh: flow cytometry data
    • Unmapped reads can create or destroy spatial autocorrelation (look at classification rates)
  • Comparing to taxonomic distributions
    • Sarah's methods
    • Patrick has Dongying's PD metric for all KOs
  • KOs versus modules
    • Katie: aggregate data for module before modeling, computing autocorrelation
    • Stephen: look at humann paper (e.g., minPath)
  • Stephen: Look at gene content and strain level variation
    • Presence / absence analysis

• Pollard lab summer intern: Kit Tse
  • Spatial autocorrelation

• Next call in two weeks: Stephen
  • Metaquery tool
  • Strain level variation