I’ve admired Prof. Ken Dill’s work on molecular modeling and protein
folding since the beginning of my PhD, so I went into his Saturday
talk “Some cell behavior is encoded in proteome physics” with pretty
high expectations. It surpassed those expectations by a wide margin!
He raised compelling arguments that the sequencing revolution calls
for reinvigorating thermodynamic approaches to modeling molecular
biophysics and protein science, along the lines of Tanford and Edsall.
Thermodynamic models never went away, of course, and structural and
molecular models offer insights not available by other methods.
Nevertheless, I think it’s significant for a pioneer in molecular
modeling to advocate biophysical models that do not place protein
structure near center stage.
To provide a little more detail, Prof. Dill’s lab has focused in
recent years on developing models that can predict protein properties
across whole genomes. His talk addressed models for two specific
influences on cell behavior, temperature and oxidation damage. In
both sections, he illustrated that models based on thermodynamic
properties (such as heat capacity) do well at explaining cell behavior
and organism development, when applied to whole proteomes.
The key point: the talk wasn’t about the proteome. It was about the
properties of proteins–not as tabulated for large sets of specific,
model proteins, but as revealed in distributions taken over proteomes.
Ken highlighted this crucial change in perspective with a disarmingly
fun metaphor. In describing his lab’s work on modeling protein
stability, he showed the formula they use, noting its simplicity and
its use of a mere handful of parameters. “With this model, your
iPhone can calculate a whole proteome in a second.”
This stands in sharp contrast to the millions of supercomputer hours
allocated via competitive grant applications, many proposing atomistic
and quantum-mechanical studies of biomolecular processes. Many of us
use these kinds of high-performance computing (HPC) facilities
extensively, just as most research fields today do. For example,
United States funding agencies such as the Department of Energy are
committed to the path to exascale computing (10^18 floating point
operations per second), which will provide simulation capabilities
vital for national security. Massive calculations provide important
insights and will continue to grow in importance and accuracy.
Despite this fantastic scale, however, understanding the effects of
climate change on the biosphere is not feasible using exhaustive molecular
modeling, and never will be.
In this respect, Prof. Dill’s talk has implications that are at least
as important as his call for a renewed emphasis on non-structure-based
thermodynamic models in molecular biophysics. Even relatively small
temperature changes can be catastrophic for organisms (as he noted, 4
K is almost irrelevant in statistical physics, but it is everything in
biology). Although large multicellular organisms have robust
mechanisms to maintain homeostasis, the story might be very different
for single-cell organisms, and thus for ecosystem response to climate