Moore Notes 7 6 11

Group Call

• Protein Db
  • Update from Guillaume
    • Re-running crashed jobs
    • PFAM clustering
      • What distance and algorithm?
      • Compare to what? Homology similarities and families
      • Maybe save for a later paper
      • Need to compare to PFAM: provides functional info and comparison of amount of clustering
      • Should also compare to COGs (have phylogenetic context)
  • Why is this db/paper different from existing protein databases?
    • Full length gene families
    • Derived from bacterial genomes
    • High-throughput, automated, easily updated with new genomes, open
  • Generation of full-length protein families and models
    • Description of workflow
    • Description of database
    • Database accessibility
  • Statistical assessment of the families
    • Family size distribution
    • Family PD distribution
    • Precision and recall distributions (local v. global)
  • The relationship between families in homology space.
    • Which families have models that recruit the same sequences?
    • Cytoscape-like network map of family homology (see attached image)
    • Clusters may represent superfamilies
  • The relationship between families in functional space
    • Hierarchical clustering of families by their pfam annotations
    • Can clusters be partitioned into broad-based functional groups?
  • The overlap between these relationships
    • Can we quantify the amount of overlap between the homology clusters and the functional clusters?
    • What does this tell us about the evolution of function across superfamilies?
  • To do
    • Make an outline (Tom, Katie)
    • Introduction (Morgan)
    • Describe workflow and metrics (Dongying, Guillaume, Jonathan)
    • Compare to PFAM
    • Compare to COGs or describe differences
    • Finish statistical analyses
    • Search vs. metagenomes and/or new genomes (compare to PFAM or COGs?)

• GBMF proposal request
  • Overhead issue
  • Katie will start outlining, Jonathan back next week