Proteins are very "sticky". Even proteins with weakly predicted interactions exhibit significant effects upon one another when all-atom simulations are performed - this is actually a major stumbling block in getting cellular simulation right.
The most complicated piece of multithreaded software yet devised by humans does not compare in complexity to the transcription, translation, and interaction events occurring in a typical human cell. The "DNA as software" metaphor is just that, a metaphor - it is an exceedingly poor model. Cellular systems are so quantitatively enormous and convoluted (yet not chaotic!) that we have to compare them to the most complicated designs our species has recently engineered just to begin to get our heads around the problem.
I haven't kept up with the state of the art in the past two years, but I remember an experiment running a ~10k node cluster for several weeks being able to successfully simulate only the cytoplasm of a cell - and 1/1000th of its overall volume at that. And the proteins were all modeled as spheres. I'm sure the art has advanced, but that is orders of magnitude away.
The most complicated piece of multithreaded software yet devised by humans does not compare in complexity to the transcription, translation, and interaction events occurring in a typical human cell. The "DNA as software" metaphor is just that, a metaphor - it is an exceedingly poor model. Cellular systems are so quantitatively enormous and convoluted (yet not chaotic!) that we have to compare them to the most complicated designs our species has recently engineered just to begin to get our heads around the problem.
I haven't kept up with the state of the art in the past two years, but I remember an experiment running a ~10k node cluster for several weeks being able to successfully simulate only the cytoplasm of a cell - and 1/1000th of its overall volume at that. And the proteins were all modeled as spheres. I'm sure the art has advanced, but that is orders of magnitude away.