BIO/CSC295 2009F, Class 11: Protein Alignments (2)

Admin:
* Math department is organizing a trip to IowaState to talk to folks
  about graduate school (biostats, engineering, etc.) week after fall
  break. Talk to Fai about getting a chance to go along!
* Neverland players doing a show this weekend (no EC, but fun)
* EC for Grinnell Singers doing an all-night vigil by Rachmaninoff
  7:30 p.m. Friday, Sebring-Lewis
* EC for Tolga's talk on Monday on complex systems (it's fun)
  Noon, 3821, LOTS OF FREE PIZZA!
* Oh no!  Family weekend is here.  Go see your colleagues' posters
* Renaisance Faire on Mac field this saturday.  Sword fighting in the rain.
* Readings for Tuesday: Chapter 5 and the HIV paper.
* One of the elementary schools in town was unexpectedly cancelled.  
  We'll have children playing in the next room.

Overview:
* Exploring Markham's data.
* Lab: Project 4.5.

Presentations (ha ha! on your projects)
* What question were you exploring?
* What data did you use (what sequences did you choose)?
* What did you learn 
  * In relation to your question
  * Other interesting things

Group one:
* Exploring the question of how to categorize progression of the disease
* Subjects 3 and 15 seemed to be abnormal rapid progressers
* Typical profile is diverse and seems to get more diverse over time
* Subjects 3 and 15 had fall in diversity
  * Not explained in the apper
* Goal: Compare rapid progressers
  * Built trees for 3, 10, and 15, using all the visits for each
  * Lots and lots of trees
* Problem: Understanding the ClustalW output scores
* In terms of these scores, 3 and 10 seemed more similar (much higher 
  average scores than 15)
* Doing more work: Could have compared to a non-progresser
* Can you change your characteristics (from rapid progressor to
  non-progressor or ...)

Group Two:
* How does diversity change over visits
  * Does diversity change at a constant rate or a faster rate over
    visits
* Data: All of the visits of one person (patient 9).
* Measure of diversity: Average alignment.  
  * Assumption: Lower scores meant more diversity.
* Didn't try to fit a curve (perhaps not enough data, and not worth
  the effort)
* Confounder: The snapshot at each visit might not include all of the
  strains

Group Three:
* Considered just one subject (subject 8)
* Compared one early visit to one late visit
* Looked at the number of mutations between stands and what those
  mutations were
  * Was it one strand that was off from the others, or multiple strands?
* In early visit, a few sites seemed to be common
  (A->G mutation was common)
* More mismatches and more random later on

Group Four:
* Worked on patient 9
* Took all the strains from 1st, 4th, and 8th visit
  * Different number of strains in each visit (5, 8, 3)
  * Hypothesis: 
    * Diversifies as pressure is put on system
    * Then specialization
  * Problem: Paper suggests that there are more strains at 8th visit
* Interesting shifts in what the mutations were
  * In the last visit, there were big insertions and deletions
* Coding tools:
  * Would be useful to have a file that has lots of FASTA entries and
    do pair-by-pair comparisons
* DON'T RELY ON SCORES ALONE!

Commonalities
* Understanding the ClustalW output scores
* Are there good ways to use the "align pairs" algorithm to align more
  than just pairs of sequences
  * Would have liked a score for the similarity between the two sequences

Questions from students
* How do you conduct a study on HIV resistance?
  * Split population into two groups; Control and experimental
  * Give placebo to control, give treatment to experimental
  * Expose everyone to HIV [No, that's not ethical]
  * Wait a few years.  
  * Count the number of people in each group that get the disease.
    (81 vs. 54)
  * Do statistical analyses.  

Questions from Sam:
* Suppose you didn't have ClustalW: 
  * How would you measure diversity?
  * How would you build it?
* One strategy: 
  * Assume everything is equal length
  * Align in the normal sense (just write one on top of each other)
  * At each position, count the diversity
  * Average or sum or ...
  * Critiques:
    * Need to take into account diversity in group
    * Need to be careful about the math - don't make the measure dependent
      on sequence length
    * Measure of diversity is affected by number of sequences
    * How do we prioritize lots of diversity at one position vs.
      less diversity at multiple positions
    * Doesn't account for position (e.g., in codon), although it looks
      like ClustalW doesn't pay attention to this either.
* Note: Perhaps we should be looking at the protein level rather than
  the mRNA level
* Question: Can we weight things according to the effects of the mutation?