Advertisement

If you have an ACS member number, please enter it here so we can link this account to your membership. (optional)

ACS values your privacy. By submitting your information, you are gaining access to C&EN and subscribing to our weekly newsletter. We use the information you provide to make your reading experience better, and we will never sell your data to third party members.

ENJOY UNLIMITED ACCES TO C&EN

Nobel Prize

‘We are in the century of the protein’

How Nobel-winning algorithms are fueling biotech today

by Rowan Walrath
October 10, 2024

 

An illustration of an algorithm predicting a protein structure, with digitized blocks on the left becoming ribbons on the right.
Credit: Ian C. Haydon/UW Institute for Protein Design
An illustration of the RoseTTAFold software predicting a protein structure.

When Schrödinger launched more than 30 years ago, computational biology was a nascent technology. Schrödinger, a pioneer in the field, relied heavily on the Protein Data Bank (PDB), a collection of solved protein structures that got its start in the 1970s.

“Getting a structure in the lab used to be—and still is—difficult,” says Karen Akinsanya, Schrödinger’s president of therapeutics R&D. Back then, a single protein structure would potentially comprise a person’s entire PhD or postdoctoral research, taking “years of work,” she says.

The entire field of structure-based drug design took a giant leap forward with the advent of AlphaFold, the protein prediction software that launched in 2020 and this week became a Nobel Prize–winning technology. AlphaFold, which is owned by Google’s DeepMind, built on the PDB and other protein sequence databases, has used a neural network to predict the structures of now millions of proteins.

Schrödinger and its ilk use AlphaFold to help dream up drug candidates with a higher degree of specificity than what was previously possible. Akinsanya says her team is using a better understanding of how proteins fold to design, for instance, small molecules that change how those proteins interact with each other. Recently, a Schrödinger team used structure predictions of a protein encoded by the human ether-à-go-go–related—or hERG—gene, to see how 14 compounds would bind with it, then design drugs that would avoid the protein since inhibiting it can elicit severe cardiotoxic side effects (Cell 2024, DOI: 10.1016/j.cell.2023.12.034).

“It’s accelerating the work of humans. There’s no doubt about that,” Akinsanya says of AlphaFold. “We are in the century of the protein.”

It’s also the century of the algorithm. Machine learning algorithms based on artificial neural nets, too, became Nobel-winning technology this week: John J. Hopfield of Princeton University and Geoffrey E. Hinton of the University of Toronto won the Nobel Prize in physics for their foundational work creating computer algorithms that can train on patterns to correctly sort data.

“Machine learning has had pervasive impacts on both medicine and chemistry over several decades,” Sarah Amalia Teichmann, a professor of medicine at the Cambridge Stem Cell Institute, says by email, highlighting AlphaFold as well as models designed to interrogate human cell data, like Geneformer and CellTypist.

“To me, it’s almost unimaginable today to be doing any drug discovery without leveraging machine learning,” Chris Dallago, a senior applied researcher in digital biology at tech company Nvidia, says in an email to C&EN.

AlphaFold and related technologies have begun making their way into pharmaceutical research. In January, Novartis became one of the first Big Pharma firms, along with Eli Lilly and Company, to partner with Isomorphic Labs, the drug discovery–focused spin-off launched in 2021 by AlphaFold coinventor Demis Hassabis. Novartis scientists are using AlphaFold to virtually screen compounds for their ability to bind to specific proteins with the goal of creating small molecules for three distinct targets.

Novartis president of biomedical research Fiona Marshall expects the partnership to cut the time it takes to turn an idea into a drug candidate by about 30%­—meaning that Novartis and Isomorphic may have small molecules ready for laboratory testing 18 to 24 months from now instead of in 3 years.

“We’ve got off to a really very impressive start,” Marshall says. “We’re very excited by how quickly we’ve been able to identify chemistry starting points for these physical targets, where we were really struggling before.”

AlphaFold coinventors Hassabis and John M. Jumper share this year’s chemistry Nobel with David Baker of the University of Washington’s Institute for Protein Design. Baker has multiple start-ups in the biotechnology world.

Daniel Silva Manzano and Alfredo Quijano Rubio joined Baker’s lab as a postdoctoral researcher and a PhD student, respectively. In 2022, they spun off a diagnostics start-up called Monod Bio with Baker. Monod uses Baker’s software to design proteins that can, for instance, create novel biosensors. Its newest product, the NovoLISA, launched last month.

“David Baker was always a great mentor and believed in developing new methods but also translating technology into the real world,” says Rubio.

Before Monod came Arzeda, cofounded in 2009 by Baker and former doctoral fellow Alexandre Zanghellini. Arzeda designs enzymes and proteins for industrial applications and has partnerships with Unilever, BP, and the makers of Gore-Tex.

For therapeutic applications of Baker’s Nobel-winning technology, there is Vilya, which was launched in 2022 to create small oligopeptide macrocycles. CEO Cyrus Harmon says Vilya is developing drug candidates for gastroenterological and immune diseases alongside cancer, with the goal of beginning human trials “in the coming years.” Baker’s software helps Vilya select for macrocycles with traits like cell membrane permeability.

And Baker’s most recent start-up showcases the biotech sector’s increasing appetite for using software in drug design and development. Xaira Therapeutics launched in April with $1 billion in venture backing—one of the largest-ever fundraising efforts among biotech start-ups—and plans to build on two of Baker’s protein prediction models, RFdiffusion and RFantibody. Cofounder and Senior Executive Hetu Kamisetty was a postdoctoral researcher in Baker’s lab over a decade ago and helped lead Meta’s artificial intelligence team before returning to therapeutics.

“David is just an incredible, incredible scientist. He’s essentially created this whole field of protein design,” Kamisetty says. “It’s about connecting the world of molecules and drugs, or chemistry, to the disease biology. . . . Protein design has essentially revolutionized that first component.”

With additional reporting by Prachi Patel and Sarah Braner

Advertisement

Article:

This article has been sent to the following recipient:

0 /1 FREE ARTICLES LEFT THIS MONTH Remaining
Chemistry matters. Join us to get the news you need.