4
Topics Lecture for today: Some philosophical/scientific background Goals of the course This really should have been lecture 1, but wasnt.
CS 482/682, Fall 2004 Lecture 4
1
Why is bioinformatics?
Note the question word. Where did this science come from, and why is there not cheminformatics or physinformatics or geoinformatics? Well, for one, there sort of is: Computational chemistry Computational physics Chaos and complexity But, with the possible exception of the third of these, theyre entirely contained within their own disciplines. Why is biology different?
CS 482/682, Fall 2004 Lecture 4 2
Why bioinformatics?
People often say that biology requires computer scientists because the data are huge. Thats not it at all. Astronomy data sets are routinely 10s of terabytes, yet they dont work with CS people outside of scientific computation Particle accelerators also produce enormous amounts of data (Terabytes per second, and Petabytes overall!) By contrast, the human genome fits on 2 CDs. All of GenBank is 42 GBases of sequence, or 10.5 GBytes of data. (as of January 2004)
CS 482/682, Fall 2004 Lecture 4 3
Why bioinformatics?
Rather, there are a few other reasons why bioinformatics exists as the interdisciplinary science it is. Practical reasons CS reasons Political reasons, etc, which I wont go into.
CS 482/682, Fall 2004 Lecture 4
4
Practical reasons
Biology, unlike physics, has undergone a complete change in the last 10 years. While biologists spend lots of time in lab, they suddenly have computers in them High-throughput methods, long common in other fields, are now found in biology Much experimentation is also in silico now Much of this makes the science quite different from what its practitioners studied.
CS 482/682, Fall 2004 Lecture 4 5
A little overblown
Actually, what I just said is a little overblown Biologists certainly have been very strong information scientists
They invented modern statistics! Genetics has always been a very mathematical science Mathematical ecology is a rich and very old discipline
And, of course, other fields certainly havent stopped:
Physics: solid state, quantum experiments, etc. Geology: nonlinear dynamics, etc.
But biology has been especially abrupt.
CS 482/682, Fall 2004 Lecture 4 6
New biology problems
These new techniques require completely new analysis techniques. Example: How much of a given protein is being produced at a given time? Ways to answer it: Northern blot (~1980): Isolate RNA from the organism, then bind to radioactively-labeled RNA for that gene.
Intensity of radioactivity = amount of transcription
Until relatively recently, that was the only answer.
CS 482/682, Fall 2004 Lecture 4 7
Nowadays
Well, first, is that really the question I wanted? I asked: how much of this protein is being produced at this time? Did I really want: Which proteins are being produced at a particular time in an organisms life cycle? Is the amount of a particular protein higher or lower at that time than usual? What else is higher or lower? Those experiments werent possible to have happen all at once before, but they are, now.
CS 482/682, Fall 2004 Lecture 4 8
Example, contd
Current techniques: Mass-spec based proteomics: Pick all of the protein being produced at that time. Break it into pieces Use a mass spectrometer to find out how much the pieces weigh Identify what they are based on matches to a database of fragment weights Not quantitative; cant tell how much is produced Can identify everything produced
CS being 482/682, Fall 2004 Lecture 4 9
Example, contd
Another technique: DNA microarrays Trap all of the RNA being produced at the time Attach it to a chip with DNA probes for each gene, tagged with a color Intensity of color = level of RNA production (more or less) (There are other techniques, but this will do for now) Both of these techniques suddenly switch to questions that feel more like computer science
CS 482/682, Fall 2004 Lecture 4 10
Before / After
In the case of the Northern blot, the major questions were: 1. Is the experiment reproducible? 2. How accurate is the measure of RNA production? 3. Is a difference between one condition and another significant? Now, these questions matter, but so do: 1. How accurate is the database lookup on the fragment weights? 2. Is there a pattern that characterizes the genes that are being turned on at the same time? 3. Whats the probability of error?
CS 482/682, Fall 2004 Lecture 4 11
End of example
Suddenly, we have computer science problems in the midst of our biology! However, this case is really a biology problem, with a CS interpretation. This is true for many cases of highthroughput biology: the problem isnt coming from CS, but rather it just feels like CS once its brought up to date.
Ironically, we wont talk about this any more in this class
CS 482/682, Fall 2004 Lecture 4
12
CS reasons
Some problems having to do with biology really just are CS problems (And, of course, thats what well spend the whole term on) The primary reason is that some of the fundamental data in biology are discrete elements. This, of course, is also true of quantum physics, but thats confusing With biology, an awful lot of the data are expressable as strings, directed graphs, vectors, or trees. And, of course, those are some of the central objects of computer science
CS 482/682, Fall 2004 Lecture 4 13
Examples
The obvious one: DNA. Well talk more about this shortly, but DNA sequences are long strings of letters over the finite alphabet ACGT. Note that this isnt sufficient, in all cases, to model DNA. DNA is, of course, a molec...