Scribe: N. Allen

Today's Handouts and Announcements

• Announce: Presentations will start next week- talk with Dr Heath before then
• Handout: Parts of Salzberg, Delcher, Fasman, and Henderson. "A Decision Tree System for Finding Genes in DNA"

Today's Topics

• Scanning DNA for coding regions
  • Intergenic DNA- Unrelated junk material and pseudogenes

  • Intergenic->ID Sequence->ATG->Exon|Intron|Exon|---|Intron|Exon->TAA->Intergenic
                                  |-------- divisible by 3 --------|
    

  • Exon->Intron boundary is a donor
  • Intron->Exon boundary is an acceptor
• MORGAN
  • System for identifying parts of a sequence
  • Uses: Markov chains, dynamic programming, and decision trees
    • A decision tree maps a finite subset of d-dimensional real space to the integers modulo two
    • Univariate if tests are of the form x_i < c_i or x_i > d_i for constants c_i and d_i
    • Oblique if tests are of the form a₁x₁ + a₂x₂ + ... + a_dx_d > c for constants a₁,a₂,...,a_d and c
    • Building a decision tree using recursive splitting
      • Let S₀ = S and T_i(S_i) = {S_2k+1, S_2k+2}
      • Criteria for splitting include balance set sizes, measures of set impurity, or the net information gain of the split
    • Pruning is replacing branches of a decision tree with a leaf containing relative probabilities of appearance
    • Partition S into sets T and P
      • T is the training set and P is the pruning set
      • Build decision tree using T and find probabilities associated with branches that could be pruned using P
  • Four Markov chains are used for finding site locations
  • Pruned decision tree is used for refining the probability estimates
  • Map donor->start, donor->acceptor, acceptor->donor, and stop->acceptor pairs using dynamic programming to find an optimal parse for the sequence

Today's Sources

• Salzberg, Searls, and Kasif: Chapter 4 and 10.