Future of Folding ... Homunculi?
Posted: Sat Jul 03, 2021 5:17 pm
If you've ever seen presentations of (~in-vivo) Human Organs-on-Chips https://vimeo.com/116674365 from Wyss Institute - it brings up interesting future prospects in-silico...
One can see Human Organs-On-Chips (a form of homunculus (homunculi?)) to be a realistic in-silico simulation target for distributed computing projects in the next 100 (even 20?!) years - with a number of cells being simulated in the hundreds of thousands to millions, rather than in the "quintillions" (the human body has an estimated 600 quintillion* proteins). There are 3,900,000 cells per square inch on the surface of human skin for comparison, so this brings things closer to reality as the number of cells needing to be simulated could be a small fraction of that. There are projects being run on Benchtop Human to simulate toxicology and the human liver already.
If we just accepted Moore's law (we all know it's breaking down, yada yada), but hypothetically speaking, within 100 years ... a future nVidia RTX 3060 level GPU could do the computational work of 34 million in today's RTX 3060s (or roughly 76 Trillion (76,000,000,000,000) PPD.
100 years isn't that remote or strange. One hundred years ago, the average middle-class family didn't live THAT much differently from families today - facing many of the same challenges. That small period of history comprising a long-lived person's lifetime, is still relatable today.
I realize this is futuristic thinking, but what does the team think about the direction of protein folding beyond the usual 5 year roadmap?
Thanks for reading - hopefully this can elicit a response, or inspire an article on the foldingathome.org web at some point. People like to see definitive end goals before contributing their time, effort and electrons to a project they understand very little, without themselves going through the same rigors of a PhD program. Even if these are hypothetical, it would give a fascinating insight into the minds and vision of the scientist behind Folding@home.
Cheers everyone.
*napkin math - total number of proteins in the human body:
- 40 million proteins in the average cell * 15 Trillion cells in the human body = 600,000,000,000,000,000,000 (600 quintillion proteins).
- It would "only" take ~8 million
future generation RTX 3060 GPUs to simulate every protein in a human body simultaneously, 100 years from now.
(of course on the millisecond timescales we're simulating, it's unrealistic to deterministically simulate how every protein in a body reacts, from the moment of injection of a new drug compound, etc.)
edit: added more data points and final request for more long-term vision and roadmaps
One can see Human Organs-On-Chips (a form of homunculus (homunculi?)) to be a realistic in-silico simulation target for distributed computing projects in the next 100 (even 20?!) years - with a number of cells being simulated in the hundreds of thousands to millions, rather than in the "quintillions" (the human body has an estimated 600 quintillion* proteins). There are 3,900,000 cells per square inch on the surface of human skin for comparison, so this brings things closer to reality as the number of cells needing to be simulated could be a small fraction of that. There are projects being run on Benchtop Human to simulate toxicology and the human liver already.
If we just accepted Moore's law (we all know it's breaking down, yada yada), but hypothetically speaking, within 100 years ... a future nVidia RTX 3060 level GPU could do the computational work of 34 million in today's RTX 3060s (or roughly 76 Trillion (76,000,000,000,000) PPD.
100 years isn't that remote or strange. One hundred years ago, the average middle-class family didn't live THAT much differently from families today - facing many of the same challenges. That small period of history comprising a long-lived person's lifetime, is still relatable today.
I realize this is futuristic thinking, but what does the team think about the direction of protein folding beyond the usual 5 year roadmap?
Thanks for reading - hopefully this can elicit a response, or inspire an article on the foldingathome.org web at some point. People like to see definitive end goals before contributing their time, effort and electrons to a project they understand very little, without themselves going through the same rigors of a PhD program. Even if these are hypothetical, it would give a fascinating insight into the minds and vision of the scientist behind Folding@home.
Cheers everyone.
*napkin math - total number of proteins in the human body:
- 40 million proteins in the average cell * 15 Trillion cells in the human body = 600,000,000,000,000,000,000 (600 quintillion proteins).
- It would "only" take ~8 million
future generation RTX 3060 GPUs to simulate every protein in a human body simultaneously, 100 years from now.(of course on the millisecond timescales we're simulating, it's unrealistic to deterministically simulate how every protein in a body reacts, from the moment of injection of a new drug compound, etc.)
edit: added more data points and final request for more long-term vision and roadmaps