Three Hundred Million Structures
America.
In Cambridge, Massachusetts, stood one of the most famous universities in the world.
Harvard University.
A school that felt deeply wronged whenever global rankings placed it second in the world.
True to its reputation, Nobel laureates among the faculty were commonplace there.
One hundred fifty Nobel Prize winners produced.
At Harvard, spotting a professor who had picked up a Nobel while still chewing over his scholarship was hardly difficult.
And in one of those Harvard biochemistry labs, the supervising professor, Professor Chris, was a Nobel laureate himself.
Chris’s achievement: the discovery of the principles of protein folding.
He was a man who had rewritten an entire chapter of every biochemistry textbook.
How much research funding could a Harvard Nobel professor pull in a year?
If he were an aging professor content to coast, tens of millions of dollars annually.
If he raised his hand and said he was going to go all out, he could haul in research grants on the order of a hundred million dollars every year.
Professor Chris was a professor with real drive.
His lab occupied two entire floors of a large building.
Add up every PhD, master’s student, technician, and administrative staff under him and the total came to seventy people.
At that scale it was less a graduate laboratory and more an independent research institute.
Chris Laboratory for Protein Structure Analysis
In one corner of the lab, the air mixed the scent of reagents with the sharp smell of metal.
Beep. Beep.
A protein PCR analyzer chirped at regular intervals.
The postdocs did not so much as register the beeping.
One postdoc, dark circles under his eyes especially deep, stared so hard at his monitor that he looked ready to burn a hole through it. Suddenly he let out a shout.
“Ooooh!”
Researchers watching their own screens nearby turned their heads.
“What? What is it?”
“I’ve got it!!! I… I think I got it!!”
A senior postdoc with ten years in the field, mid-pipette, came over.
“For real? Let me see.”
The dark-circled postdoc who had shouted clasped both hands together and prayed.
“Oh Jesus, please, let it be right…”
Senior postdocs nearby gathered to verify the claim.
One hour of checking. Two hours…
One of the seniors spoke quietly.
“Here… one amino acid… doesn’t match.”
Blink. Blink.
The dark-circled postdoc only blinked for five full seconds.
Then he started clutching at his hair.
“Aaaargh…”
He rolled across the floor.
The postdocs with ten-plus years of experience clicked their tongues as they watched him thrash around.
“Tsk tsk.”
“That’s just how it goes.”
“That patch of floor’s dirty. Go roll over there instead.”
Fitting a protein’s 3D structure was, in practice, solving a puzzle.
It was the process of building a three-dimensional model from the shadow-like two-dimensional data produced by X-ray diffraction.
The difficulty level of assembling a ten-thousand-piece 3D puzzle while looking only at its shadow.
In this lab, the number of protein structures solved in a year was roughly one.
Cases like this—almost there, then not—were painfully common.
The postdoc whose dark circles had only grown denser sat slumped on the floor, hollowed out.
If he had nailed it, he could have put his name as first author on a top-tier paper.
Which meant he could have landed a faculty position at a respectable university.
The ticket to a professorship had flickered past his eyes like a mirage.
He spoke in a vacant voice.
“Haah… seniors, is this even work for a human being?”
The senior in the next seat, whose hair had already retreated dramatically, nodded.
“Yeah. Not work for a human being.”
The dark-circled junior stared blankly at him.
The look said: if it isn’t work for a human being, why are we still doing this?
“But you know perfectly well what happens when we actually solve one of these structures.”
“Haah. Knowing is one thing.”
“Solve a single enzyme structure and suddenly inhibitors for that enzyme, activators, mutation checks, cancer-variant analysis, vaccines, anticancer drugs, antibiotics—all of it… becomes possible.”
It wasn’t as if the dark-circled postdoc didn’t know that.
The postdoc senior patted him on the back.
“Your dark circles have dropped even lower today. Still, the further those circles drop, the more lives you’ll be able to save.”
Protein-structure puzzle-solving of insane difficulty.
It was extraordinarily hard work, yet every researcher in the room understood that this work saved lives.
The dark-circled postdoc finally allowed himself a grudging complaint.
“I know. I do. But I’ve been trying to solve this one for six months…”
The senior answered as if he already knew.
“It’s fine. What’s life, right? Just start over from the beginning.”
Just then a postdoc with fifteen years after his doctorate walked past and tossed out:
“I’ve been at this for fifteen years and I still haven’t solved a single one…”
The fifteen-year postdoc.
Back in his Harvard undergrad days he had been called a genius.
IQ of 160, Mensa member.
Even that genius senior, after master’s, doctorate, and fifteen years of postdoctoral work, had never correctly solved a single protein structure.
The dark-circled postdoc looked at the fifteen-year man.
Salt-and-pepper hair already showing.
A bleak thought flickered through him: is that my future?
Then—
Bang!
The lab door flew open.
Every head in the room snapped up.
A master’s student who had opened the door shouted without even catching his breath.
“Holy shit! Holy shit!!”
“What? Did something blow up?”
He thrust the tablet in his hand high into the air.
An icon on the screen was blinking.
“What kind of title is that?”
“I’m serious, this is insane. Go in and look. I’ll send the link. Or just search JH Research protein structure and go in that way.”
He was still panting, looking deadly serious.
A few people started searching online.
“JH, protein structure?”
“What, did they figure out a few structures or something?”
They searched without much expectation.
Protein 3D Structures: 300 Million
The researchers looked at the site’s title and briefly experienced brain freeze.
Three hundred million protein 3D structures?
“Where the hell did they pull that bullshit from…”
They kept clicking.
—Protein search
—3D model
—Drug recommendation
Click.
Click.
For a while the only sound in the lab was the clicking of mice.
Scroll. Scroll.
The researchers rotated the three-dimensional protein structures on their screens.
“They made it look pretty good.”
“There’s no way this is real, right?”
They entered the sequences of the protein enzymes they themselves had been studying.
M—A—S—D……
A—S—E—F……
“It looks plausible, at least.”
The dark-circled postdoc pulled up the structure of the protein he had been working on.
There it was: the 3D shape of the protein he had been trying so hard to solve.
Coils, zigzags, a labyrinth of tangled folds.
It was roughly similar to the model he himself had built.
Scroll. Scroll.
The 3D model could be zoomed, shrunk, rotated in every direction.
The dark-circled postdoc compared JH Research’s stereo model against the data he already possessed.
Months of X-ray diffraction images he had taken.
Building a structure from diffraction data was hard; verifying whether a structure was correct against that same data was fast.
“…Huh?”
The data lined up perfectly.
“There’s no way…”
He compared sheet after sheet.
One. Two. Three……
Everything matched.
“This is actually right?”
The words rose from every corner of the room.
“The one I was analyzing is right too.”
“Mine as well…”
“Holy shit… why is this correct?”
“No, but it says three hundred million. Three hundred million? Does that even make sense?”
They were, by any measure, the top laboratory in the world in this field.
Even in this lab they managed about one structure a year.
The total number of independent protein stereostructures solved by every laboratory on Earth so far was roughly a thousand.
Of course, if you counted every variant that differed by a single atom at the molecular tip as a separate structure you could inflate the number to a hundred thousand, but the number of truly independent proteins solved was only around a thousand.
Decades of work by the entire planet: one thousand.
Even if this site had counted every single-atom tip variant, it still meant they had solved at least several million.
An unbelievable website.
“Is this… a dream?”
“Hey… why are you in my dream?”
“What the hell…”
“Dude, this has every single thing our lab has ever worked on.”
The dark-circled postdoc whispered at the screen.
“But how did they even figure all this out?”
A site that should have been impossible.
They soon discovered the method.
“They fed existing data into an AI, the AI found the regularities, and then it just… analyzed every structure?”
This was Harvard.
Who did they think they were?
Of course they had done computer modeling themselves.
Several researchers in the room had written their entire doctoral theses on computer modeling.
And yet here they were.
“The AI did all of it?”
“It really wasn’t work for a human being, I guess.”
“An AI can do this?”
“What kind of monster did JH Research build…”
The dark-circled postdoc shouted.
“Wait a second! We might have been hacked.”
“What are you talking about?”
“They hacked the data we were working on and only uploaded the structures that match ours.”
“And everything else is fake?”
“Three hundred million is ridiculous. Professor Chris is probably pranking us with a surprise.”
“Yeah. Is today April Fools’? That explanation’s more convincing.”
Reality denial.
Most of them were simply vacant.
A hidden-camera prank felt more persuasive than the idea that three hundred million structures had all been released.
One hour.
Two hours.
Ten hours.
…
No matter how many ways they verified it, the only result that came back was that it was correct.
An eight-year postdoc finally pulled himself together and asked the fifteen-year man beside him.
“Senior.”
“What?”
“In your eyes, the JH site looks real, doesn’t it?”
“Yeah, it looks that way.”
“But we decided to dedicate our lives to solving protein structures and saving humanity from disease.”
“We did.”
The postdocs stared blankly at JH Research’s website.
“So… what do we do now?”
JH Research’s protein-structure site hit the relevant industries like a tidal wave.
“Professor Chris, is it true that three hundred million protein structures are up on a website?”
“Ah, Professor Martin. Congratulations on this year’s Nobel. And that protein site—ha, ahem, yes, it’s real. Our entire lab piled onto the verification and everything checked out.”
“How in the world can an AI do that?”
“JH Research once opened their AI to the general public. I went and talked separately to people who used it then. They said it was smarter than a person.”
One professor spoke in a trembling voice.
“Hah… until now every researcher on the planet combined had only solved about a thousand structures…”
Another professor, looking at the site himself, muttered.
“But this is… three hundred million? They didn’t just inflate the data, did they?”
“We went through it thoroughly. That wasn’t it. Literally three hundred million independent protein structures.”
“Hah… let’s see. 3D models? Sequence–structure–even the molecules that bind easily to each region?”
“This is going to rewrite the molecular biology textbooks themselves.”
“Why bother rewriting textbooks? Just use the site as the textbook.”
A quiet elderly professor spoke up.
He too was a Nobel laureate.
“No matter how you look at it, this is Nobel material.”
For a moment the conference room went still.
No one could deny the statement.
“You’re saying we should submit a nomination right now?”
“Yes. There is no precedent for an achievement like this. It’s on a scale that shakes the entire paradigm of the life sciences worldwide. This one deserves to be given. I’ll put in a phone call myself.”
Stockholm, Sweden.
The meeting room of the Nobel Committee.
The recording officer quietly laid a document in front of the Nobel prize examiners.
Summary of the public release of approximately three hundred million independent protein 3D structures, generated by artificial intelligence trained on existing experimental data, with reported concordance of 99.999% against experimentally solved structures.
The chair of the Nobel examining committee adjusted his glasses and spoke.
“Korea… claims to have solved three hundred million protein structures?”
The specialist in molecular biology beside him tilted his head.
“Is it genuine? They say it’s AI prediction—what about the accuracy?”
Another committee member who had been scanning the documents answered.
“Accuracy is reported at 99.999 percent concordance when compared against existing experimentally based structures. There are even verification cases down to the positions of OH groups at electric-unit binding sites.”
“OH groups… isn’t that the part that drives experimentalists to despair the most?”
“Graduate students have cried a great deal over that part, yes.”
“And right now papers keep appearing that verify whether JH Research’s site is correct.”
“The research direction has already shifted. People are no longer trying to solve protein structures themselves; they are trying to confirm that the content of the JH site is accurate.”
“Has the site been found to contain errors?”
“None so far, apparently.”
From one side of the room a committee member with a kindly face spoke slowly.
“The real thing appears to be the real thing.”
The chair closed the file with a soft thump and said quietly,
“This is not simply a matter of having solved protein structures. This is a contribution that has changed the future of the life sciences.”
From the next seat another member asked carefully,
“But if the research was done by AI, should the Nobel go to the AI, or to the person who developed the AI?”
“Obviously to the person. The principle is that Nobel Prizes are awarded only to the living.”
Nobels were not given to the dead.
Having one’s achievement recognized while still alive was therefore one of the difficulties of winning one.
“Presumably the Nobel Prize in Chemistry would be the appropriate fit?”
“Is the representative a chemist?”
“He is not.”
“We give a chemistry prize to someone who is not a chemist?”
“Then the medicine prize?”
“Is he a medical or pharmaceutical doctor?”
“He is not.”
At that, one person raised a hand.
“I looked into it. The head of JH Research is a high-school graduate.”
“High-school graduate?”
“Yes. He only completed high school.”
“Really? We are to give a Nobel to a high-school graduate?”
“Achievement matters more than degrees, of course.”
“But… under what category do we award it? It is neither chemistry nor medicine.”
“A Nobel Prize in Computing?”
The chamber fell quiet for a moment.
Then someone snorted a laugh.
“There is no such prize.”
Originally there had been only five Nobel categories.
Physics, Chemistry, Physiology or Medicine, Literature, Peace.
Economics had been added in 1968.
Someone muttered in a low voice,
“Then shall we create a new category for the occasion? They created Economics, so why not? Something like a Nobel Prize in Computing.”
“You’re joking?”
“Half in jest. This achievement does not fit neatly into any existing prize, yet it is certain that the work is a great achievement for humanity.”
Another committee member sighed and said,
“That is exactly why we are troubled. If AI-predicted protein structures have advanced global new-drug development by ten years… is that chemistry? Medicine? Or do we have to invent an entirely new Nobel?”
Silence settled over the chamber again.
Someone opened his mouth slowly.
“But one thing is certain.”
He closed the document.
“This is an achievement that will remain in history. Whatever decision we reach, the world will be watching this research.”