This page is part of an in-progress commentary by Roger
Frye
on High Altitude
Thinking: The International Informatics Summit.
Start with short presentation by each panelist.
1) Ajay Royyuru
Shows pyramid of organization of info in biology. Ecosystem at top, down to comparative genomics, pharmacologists, etc.
Patterns in Protein Sequences. (I can't read.) Looks like a table of patters and their meanings -- adictionary.
Structural biology: from sequence to structure. Chart of protein pattern pictures. He calls them bets comparing to Casti's football bets.
Protein sequence and applications on Blue Gene (the IBM computer).
Functional Genomics and Systems Biology. Picture of cycle experiments -> data mining -> reverse engineering -> theoretical issues -> simulation -> back to experiments.
2) Annette Adler, Project Manager, Life Science Informatics, Agilent
Challenges of pharmaceutical discovery: competitive pressures, opportunies yet challenges from emerging genomics and proteomics, data overload, gap between data and biological understanding, collaborative obstacles.
Studies of life scientists (she was trained as anthropologist.): in situ studies of people and technologies in research and pharmaceutical discover, qualitative studies of work, contextual understanding of work.
Informatics in pharmaceutical discovery. Picture of gene expression or proteomic expression -> treatment and instrument control -> data analysis
Agilent's life science informatics. Bridging informatics (data types, collaborators, models, time). Pertnership with Aglent's Life Science and Chemical Analysis Group (shared formative vision for both research and business, successive generations of research support and business growth, symposia)
3) Bassil Dahiyat, President and CEO, Xencor
Does information help drug discovery? Jury still out, but believe.
Pharmaceutical R&D spending and output. Linear growth to 1991. 13% aggregate growth rate. But stalled. Was a peak in 1996, but a congressional anomaly. Quality of discoveries going down because of the race.
Lost in the mix. chart shows mix of drug like moleules (10^200) and genes (35000).
Xencor hypothesis. The understanding and rational manipulation of proteins not only can create better drugs. It is the BEST way to create better drugs.
Iformation challenge: evolutions limits. Table of length N, M aa=> M^N aa sequences
3-> 10^4, 5->10^5, 50 -> 10^60 = mass of the earth.
Protein design automation. (didn't understand)
Making drugs: Using information is expensive.
Protein design in practice. Animationn shows picture of molecul and their intervention.
New protein drug testing. Rescue of human U937 cells from TNF-alpha induced expression.
PANEL
Erika asks does and how long will it take to find new drugs with infomatics? Ajay says yes. Big change since his grad school. Acquiring data is easy. No substitute for making sense. Bassil agrees -- big difficult problem. Question from audience: is the holy grail really the ultimate drug model. How can non IT people do this? Answer from Bassil: Are IT intelligent. Ajay says grail is to understand how life systems operate. Then can use that knowledge to make drugs. Can manufacture molecules, but not a bacterium ( a 480 body problem). Annette says you can give people tools. Won't get there by jumping to conclusions.
Erika as a journalist: hype that will be able to do all of our experiments in silico. Anjay says that understanding will obsolete in silico experiments. Have to co-evolve.
Question: relating to Casti's talk about development of probability theory, but do not yet have the math for complexity. the trough of disilusionment with hit biotech. Not a 480 body problem because there is a scale free power law that works. Ajay answers don't even know rates of 480 bodies.
Question from Stu Kauffman. Like slide on experimental design and reverse engineering. We know there is a gene regulatory network. Not just on and off. Have to reckon with the fact that we are taking on a problem that is harder than any before. Lists many factors. Barely have a conceptual framework. Working with Carnegie Mellon using Bayesian analysis to reverse engineer. Need invention of new experimental procedures. What is the avalanche of effects -- turns out to be a power law by the way.
Erika asks what are the areas where bioinformatics will hel? Bassil says positive effect in lab looking at large avalanches. Need to look at phenotypic responses once in the hands of doctors. Bassil says that when have lots of data, easier to do correlation studies. Starting to draw boundaries on the size of interactions. Annette says tech mining has not yet come up. What does my colleague know? Visualization not only of molecules, but of multiple kinds of data, incomplete, dirty.
END