Proteins are not static entities – since we live at about 300 degrees above absolute zero there is constant Brownian motion. However, looking at deposited X-Ray structures, one might get the impression that the structures are rigidly sitting in vacuum – nothing could be further from the truth! I like the analogy with early photography :
because the photoplates were not that sensitive, long exposure times were necessary. Hence people had to hold very, very still for several minutes in order to get a decent picture. Photographers had special setups and chairs with neckbraces to keep the poor subject in place. This apparatus is the analogy to a protein crystal – it keeps the proteins in place, floppy and moving parts will not show up on the resulting electron-density maps.
The photographs of our great-grandfathers leave us with the impression that they were very stiff people, largely devoid of any humour. That’s probably not true, but how happy and lively would you look if you had to sit still for quite some time in your best outfit with your head squeezed onto some weird mechanical contraption? The same holds true for proteins.
In order to get a more “lively” dynamic picture one could run either large-scale molecular dynamics simulations for several months on a hu-Hu-HUUGE cluster or use methods based on anisotropic network models such as CONCOORD, ProDY, or this ANM server instead.
As Bosco put it
“Minutes of calculation with CONCOORD (constraint network) produces pretty much the same result as hundred of thousands of hours of CPU time with DESMOND (Molecular Dynamics). There’s a lesson in there somewhere.“. To me that clearly states that for coarse-grained, long-term dynamics, network based simulations are the method of choice.
Similarly, Ding & Dokholyan estimate that DMD (Discrete Molecular Dynamics)
“…compared with traditional MD (…) can be 10^8–10^10-times faster.” (Ding and Dokholyan. “Simple but predictive protein models“. Trends in Biotechnology (2005) vol. 23 (9) pp. 450-5).
That’s the same orders of magnitude as accelerating from walking or running (1-10 m/s) to light speed (ca. 3*10^8 m/s)! Kind of similar and presented recently at the last RECOMB : “MIT’s Ensemble Approach Speeds Protein Folding with Minor Compromise in Accuracy”
Back to “real-world” analogies: I’ve always imagined the interior of a cell like a bustling marketplace – in Freiburg it is possible to take the stairs up the 116m high tower of the Münster and see the semi-chaotic movement of people shopping for fruit&veg on the Münsterplatz below. One starts wondering how people actually get what they want and find their way while constantly bumping into one another, sometimes meeting friends and stopping for a chat.
Indeed, that view is not far off this dynamic picture of the interior of the cell: The Elcock Lab has done large scale simulations and made them available as movies, in addition there are detailed ftp instructions for downloading the full quicktime file (>300MB).
The original publication is “McGuffee and Elcock. Diffusion, crowding & protein stability in a dynamic molecular model of the bacterial cytoplasm”. PLoS Comput Biol (2010) vol. 6 (3) pp. e1000694 and there are 3 very positive evaluations on the Faculty of 1000 (F1000) site.
Since DMD already found an application in investigating the Macromolecular Crowding Effects on Protein Folding – Wouldn’t it be great to see such large-scale simulations combined with fast calculations on the flexibility of each molecule? If you see something useful in that direction, keep me posted!
(Thanks to Henning Stehr for hints & links by email)