by Paige Donner
Proteomics - the study of proteins
DeepMind - Google’s AI Division, published the entire AlphaFold2 database as open source in 2022
AlphaFold - ‘can accurately predict 3D models of protein structures (over 200M of them) and is accelerating research in nearly every field of biology’ LINK
ProteinMPNN - developed by University of Washington, AI based tool that gives scientists the ability to reverse engineer protein structures GitHub
DeSci - decentralized science
Science DAOs - science focused decentralized autonomous organizations, the majority of which are now bio-science oriented (but don’t have to be)
Blockchain - immutable, distributed ledger technology
#DeSci DAOs are blossoming just as AI and ML open source databases and systems are unleashing the ability for scientists and engineers to conduct innovative research at breakneck speeds.
At this time last year there existed 4 science-focused DAOs. VitaDAO, LabDAO and PsyDAO were all spun out of Molecule.io. They are all bio-science focused DAOs and are, arguably, the first to enshrine the concepts of #DeSci, or decentralized science, into a DAO structure.
And then there was us, FrontierDAO, established in September 2021. We came along with a science focus, too. Though our science focus is not on bio-sciences but rather on fusion energy innovation, space exploration technologies and climate solutions with an emphasis on innovation, Test & Evaluation and, ultimately, commercialization.
This article focuses mainly on how major breakthroughs in AI, and their release as open source, are fueling this massive upswing in science DAOs, particularly bio-science DAOs, and #DeSci. At last count - Sept. 19, 2022 - there were 25 bio-science focused DAOs active in the #DeSci movement. While there are only 3 DAOs focused on space exploration, 1 on climate and environment and 1 on fusion energy.
So what is it that these bio-science DAOs are seeing that the rest of us are not?
Well, in a word, AI.
Just a few months ago, AlphaFold 2 AI predicted, accurately, the structure of over 200 million proteins that exist on this planet. That represents nearly every known protein structure on Earth. Once DeepMind published that database open source in its entirety, it was all systems go for scientists and researchers around the globe.
How AI and ML are Ushering in #DeSci - Breakthroughs at Breakneck Speed
And this is the point that I have heard little open discussion of in the DeSci Web3community.
But it is the proverbial elephant in the room.
It is this: Now that AI (artificial intelligence) and ML (machine learning) techniques can be readily used in research and laboratories, we are sitting at the dawn of massive breakthroughs in scientific and engineering research. Moreover, these breakthroughs will be coming at breakneck speeds.
The Immortal Jellyfish
Let's examine, for example, some recent research coming out of the University of Oviedo in Spain on the jellyfish species, Turritopsis Dohrnii. It has now been dubbed the ‘Immortal Jellyfish’ because of its capacity to regenerate itself. For researchers in Longevity Science, this represents a true breakthrough. By the way, the first science DAOs initially focused exclusively on longevity research, funded indirectly by the Ethereum Foundation, via the Methuselah Foundation.
Recently, DeepMind was used to sequence the proteins of Turritopsis Dohrnii. The researchers were able to identify variants in the jellyfish’s DNA that allow for this immortality. In essence the jellyfish ages from young to old, reproduces, and then reverse ages back to young again, all the while remaining the same genetically. It achieves this by preventing its telomeres from shrinking. (*Shrinking telomeres is what triggers aging in an ultra-simplified explanation of the aging process.)
But this jellyfish has genes that enable it to protect the ends of its telomeres. In this way, it ‘cheats’ death.
Here’s the long and short of it:
“Now that we understand the mechanism that enables these jellyfish to shirk a “normal” death, researchers can analyze similar genetic sequences in humans. We may be able to design a therapy that stops our own telomeres from shrinking… or at least slows the process.
The researchers here used DeepMind’s AlphaFold 2 when conducting their analysis of proteins.” - Brownstone Research
So, within just a few short weeks of AlphaFold2 being published by DeepMind as open source data, scientists and a team of researchers have already achieved a major breakthrough in longevity research.
When I referenced this breakthrough on a recent Twitter Spaces (Sept.19th) one of the bio-sciences researchers asked if the research had come out of UC Berkeley or University of Washington. I replied neither. My (accurate) answer that it came from the Universidad de Oviedo in Spain was met with skepticism by my peers, most of whom were U.S., Canada and UK based scientists and researchers, fellow #DeSci DAOers building in the Web3 space.
Hopefully the example of this immortal jellyfish now makes it clear to the reader how and why AI and ML are now ushering in the age of decentralized science? This movement is fondly referred to as #DeSci by those working and building in the space. In this context, decentralized refers to the decentralization of power, knowledge and resources.
Just as the art world is being taken by storm by the likes of DALLE-2, Midjourney and Stable Diffusion, so is the science world being jettisoned into a new era of research and development with AI. We are today at Ground Zero of the explosion in proteomics that is destined to unfold from this point forward.
Years down the road, we’re going to look back on 2021/2022 as a time when AI hit an inflection point in medical science and biotechnology… and led to a remarkable number of scientific breakthroughs and therapeutic developments.
- Jeff Brown, Graduate of Purdue University
University of Washington’s MPNN
Building upon wins, a team of researchers at the University of Washington in the United States just released an AI tool called MPNN. It’s a reverse engineering tool that allows scientists to, yes, reverse engineer protein structures in the lab. It even identifies the protein’s amino acid sequence. Essentially this gives scientists an ‘instruction manual’ they can use to recreate that protein in their lab.
The implications for both synthetic biology and therapeutic developments are not to be underestimated. In essence, scientists are now ‘empowered to design new proteins based on their structure alone, then use ProteinMPNN to determine how those proteins can be manufactured.’
What are some potential applications? For therapeutics, cures for rare diseases are now within our sights. And applications for synthetic biology are just as riveting - what about bacteria that can eat CO2 emissions?
Both Alpha Fold 2 & ProteinMPNN are published as open source. Anyone can use them.
And many bio-science focused DAOs already are and/or are poised to do just that. In fact, one of the first #DeSci DAOs to be established just got a nibble and a $500K investment from Big Pharma, Pfizer to be exact.
Perhaps this is why #DeSci went from being ‘utterly fringe’ this time last year, to sharing double-billing on an MIT-hosted conference in September 2022? (#DeSci Boston)
AI is at the root of this decentralized science (R)evolution. There is no doubt about it. And AI, coupled with ML, will be doing similar heavy lifting for engineering research teams now and well into the near future.
Sources and Further Reading:
Science, The Endless Frontier by Vannevar Bush
DAOs Are Not Corporations, by Vitalik Buterin
The Bleeding Edge, Brownstone Research
What Is #DeSci - Decentralized Science, Paige Donner
Jumpstarting America, Gruber & Johnson
Protein MPNN, Github
Timeline of A Breakthrough, DeepMind Alpha Fold