Saturday, January 29, 2022

Enzyme Pipeline


The next twenty years are going to make changes in the first 60 years of our experience look quaint. Coming fast are AI, Robotics, IOT, radical advances in materials science and engineering, Biotech, movement into outer space (Starship) as well as innerspace (VR), CAD, networking and computing, big data, etc. It's not just the incredible advances in things individually, it is the synergy of those advances. A perfect example is something that made my friend a lot of money investing in Nvidia. Advances in monitors, CPUs, network buses, and software led to graphics cards as separate engines to run highly parallel graphics routines. Techniques enabled by advances in AI made GPUs into GPGPUs and, for instance, predicting protein folding, and beating everybody/anything anywhere in Chess followed. Protein folding is a biggy, once we can design and build the right enzymes we can re-engineer bacterial genomes to build enzyme pipelines that would do chemistry impossible otherwise. There are some useful chemical structures (explosives for instance) for which we have designs we know work because they have an activation barrier on the reverse side that renders them stable. We don't know how, or even if it is possible using conventional chemistry, but get this, I should probably find somebody to patent this:


It does not matter how you house the pipeline. I would think that it would just be easiest to house it in bacteria because the engine that builds the device and the engine that builds the engine that builds the device requires such an environment. That's likely a patentable idea itself, but there is something more:

You design and build a pipeline that takes precursor molecules, and alters them step-wise such that multiple reactions assemble something that would otherwise be essentially impossible. 

1) Precursor molecules floating about
2) Precursors bind to enzyme activation site
3) Bound precursors Red and Blue start self assemble
4) Bound precursors Red and Blue self assemble
5) Assembled Red and Blue move across second enzyme
6) Red/Blue begins self assembly with Purple
7) Final output molecule finished
8) Finished molecule released



 
Believe it or not, that's not the big trick. The big trick is this: 
You find/design a molecule you want made
You give it to a chemical building device 
Software reverse engineers intermediate steps needed
Software uses AI to figure out what proteins will fold to enzymes that will do the job
Software determines what intermediates are needed to build the pipeline
Device creates DNA that codes for the proteins and enzymes
Device builds pipelines
Device runs pipelines
Device collects and delivers released molecules. 
Fin.

A few enormous breakthroughs are required for the above, and as far as I know they have been made and are being refined for production now. 

The pipeline above has a lot of fussy bits, but it is buildable. When we were in high school it was not even conceivable that something like this was possible. Now it's conceivable and in a decade or two it will just be technology. 

A set of pipelines similar to the above might be able to assemble carbon nanotubes to make cable for a space elevator. It could also be used to productionize the building of these pipelines so the thing can bootstrap upward. 

No comments:

AI Alignment and Security Now!

As far as I know, people charged with 'Alignment' at AI companies are not convinced things are safe enough and some have quit becaus...