AT A GLANCE
PhD in biochemistry and molecular biology from Harvard University
Before co-founding the Molecular Sciences Institute with Sydney Brenner in 1998, served as associate professor at Harvard Medical School and Massachusetts General Hospital.
Recently named CEO, president, and chair of MSIs board of trustees.
QThe Molecular Sciences Institutes stated mission is to develop tools for predictive biology. How far along toward that goal do you currently stand?
AI think the people at the institute have pretty much sussed out the direction that the computer and mathematical methods development needs to proceed in. The strategy and tactics to be employed are reasonably clear now. In terms of building quantitative models of intracellular biological systems, theres been some significant progress, particularly led by Larry Lok, our math guy.
QWhat challenges do you face in this work?
AThe main challenge in the part of the work at the institute thats under my direct responsibility is to bring into being experimental methods that will produce the kind of information needed for quantitative models like this.
Were attempting to couple dry and wet work, and the wet work proceeds at its own frequently slow and frustrating pace. Were in the trenches on a number of things.
QHow many people does MSI have working on the wet lab side and how many are on the computational side.
ATheres maybe 10 people in the lab and three people doing simulation work.
QWhat is the workflow? Does experimental data get pushed onto the people doing simulation?
AIts much more interactive. The wet lab people are talking to the computer people all the time about whats possible and what isnt.
Because of the relative difficulty of doing experiments in the lab in the real world of biology, it would not surprise me at all if eventually some of the simulation methods were employing were different than the ones were now pursuing because theyre a better match to the kind of experimental data that we can get. Plans may change.
QSo the methods that you have so far are based on data you already have, but as you gather more data the methods will have to evolve?
AThe methods we have for simulating things are based on data that we have or on data that we think we ought to be able to get. But if we develop ways to measure other things, for example the probability that a given molecular entity at a given time slice changes into some other molecular entity, that would change the way we do the simulation, and depending on what sorts of ways we can devise to get that information, it will change.
QSo if the bottleneck is still the experimental process, how do you propose to overcome that?
AWere working away at new methods. Thats fun. Theres almost no concept from applied physics that Im not willing to plunder, so this is giving me an extremely broad and rather shallow familiarity with all kinds of physics stuff.
For example, one way to measure the rates at which some of the reactions that we care about occur that ought to be doable is to examine directly the masses of the proteins in cells that are, say, participating in the signal transduction pathway.
If we could learn to identify by mass the individual protein players and then their changes in mass as theyre phosphorylated, you could track as a function of time the rate of phosphorylation through the system.
QHow far along are you on building these simulations?
ABuilding a simulation is kind of like building a cathedral. People will work on it in a spurt and then theyll wander off and do other things for a year or two and then come back and work on it some more. In all likelihood it will never be a finished thing. Its something that will hopefully give you quantitative insight into the behavior of a particular pathway.
[With] the knowledge we need to gain and having the simulation co-evolve with that I anticipate being at this for a decade.