Published: January 4, 2025
Think of a problem in science. Any problem: O-shot drug discovery. The virtual cell. Perfect disease prediction. Room-temperature superconductors.
Whatever it is you were thinking about, I'd like to convince you that it would be solvable if you had a machine at your disposal capable of doing an arbitrarily accurate MD simulation of an arbitrarily large system (in finite time).
If you could simulate an entire cell, the drug, and possible even the entire organism of a patient, if you care about drug delivery (or any ADMET property for that matter), you would be able to perfectly predict if that drug would work or not. You could see at any desired level of detail what the phenotypic effect of the drug on the cell is. Did this protein increase in concentration? Did the growth of the cell normalize? Did the cell survive or die? And so on. Given you the MD simulation is infinitely fast you could do this for an arbitrarily large library of molecules (you could screen the entire chemical universe) and pick out the globally optimal chemical compound to cure this disease, no side-effects, perfect delivery as an orally bioavailable pill.
The same is evidently true for disease prediction, just simulate the entire body and everything in it at perfect detail. Same is true for screening every conceivable combination of materials and measuring the room-temperature conductivity in silico. =
Now this blog resulted from an argument I had with my co-author where I felt insulted (as a chemist) by the insinuation that everything chemical problem under the sun is just trying to cleverly avoid using MD by finding a more efficient (i.e. cheaper) alternative method of arriving at the answer. I was insulted - but wrong. An infitely accurate, infinitely fast MD simulator solves everything. But so it should and it's really not at all surprising that it does.
Now this gedankenexperiment is not quite as void and circular as saying that an all-knowing oracle has the answer to everything. (Which is true by construction.) But it is pretty close to the short fable by Jorge Luis Borges On Exactitude in Science, where the map of an empire grows so large and becomes so accurate that it takes on the size of the empire itself. A simulation a map scale of 1:1.
A model that has become so large it ceases to be one in the first place. It rejects the most fundamental premise. A model is a simplified version of reality. An infinitely accurate, inifintely fast MD simulation is just rebuilding every atom as bit and would take in fact more energy and matter than the universe it is trying to simulate itself.
← Back to main.