Frozen as lattice-like crystals, the two proteins look almost identical. For scientists trying to design a drug that locks onto one but not the other, their resemblance is a significant problem.
Relay Therapeutics, a five-year-old biotech based in Cambridge, Massachusetts, has raised nearly $1 billion on the promise that the company has the right tools to solve it.
Armed with powerful computing and a kit of newer drug-hunting techniques, Relay can work out how proteins move and change shape. Sanjiv Patel, the company's chief executive, believes these molecular movies can help Relay tell similar proteins apart and, as a result, design better drugs in less time.
This year, Relay pulled back the curtain on what it's been building, revealing its first three drugs under development. Each targets a type of cancer and is a step forward in Relay's attempt to continue a drugmaking journey that began in 1989 with one of biotech's pioneers, Vertex Pharmaceuticals.
"We're the fruition of what Vertex wanted to create 30 years ago," Patel said in an interview.
Vertex, and then others, set out to make more potent drugs by understanding the three-dimensional structure of disease-causing proteins. Knowing the structure of the lock, they reasoned, leads to a better designed key. This ‘rational' approach was an advance from earlier, less precise methods.
Missing from the images they constructed, however, was a detailed sense of how proteins move. Never still in their natural state, proteins flop, twist, curl and fold — less a lock and more shape-shifting putty.
"Everyone in the structure-based drug discovery field has long recognized their targets weren't static," said Joshua Boger, Vertex's founder and its CEO for two decades. "Whatever method was used to try to understand a biological process or design a drug that treated a protein as a fixed block, was making a drastic assumption."
Relay's pitch is that it doesn't have to assume. Formed by a founding team that included longtime Vertex scientist Mark Murcko and billionaire hedge fund manager David Shaw, Relay appears to have already won over Wall Street. The company went public this July, raising $425 million in one of biotech's larger initial public offerings this year. On the first day of trading, shares shot up in price by 75% and have remained there since.
While investors may be sold, proving that protein motion can lead to superior medicines will be much harder.
"Protein motion could be important, but there are a lot of other things that could be equally or more important," said Derek Lowe, a longtime drug discovery researcher, in an interview. "How do you know that this is the knife you need to cut open this problem?"
'Like looking at photographs of identical twins'
The two seemingly identical proteins, Relay discovered, don't move the same way. A flap on one is looser, and extends outwards from the tangled mass of amino acids more often than on the other.
Both are members of a family of proteins known as fibroblast growth receptor factors. Genetic alterations in one, called FGFR2, are thought to play a role in a wide range of cancers, making it a target of choice for drugmakers.

Blocking just FGFR2 is hard to do, however. Two recently approved anti-FGFR drugs, Incyte's Pemazyre and Johnson & Johnson's Balversa, also inhibit the activity of FGFR1, which can lead to elevated levels of phosphate in the blood. As a result, patients on both drugs have to be watched carefully and sometimes need to be switched to lower doses, which Relay argues limits the drugs' effectiveness.
"If you design a drug that inhibits the whole family, it causes a lot of toxicity," said Don Bergstrom, Relay's head of R&D, in an interview. "No one has really been able to leverage selective drugs, because it's like looking at photographs of identical twins."
Photographs like these are the starting point for most drugmakers, Relay included. Typically created by bombarding crystallized protein with X-rays, the images can reveal "pockets" within the protein structure that researchers then target with chemical compounds.
At Relay, researchers feed these snapshots into computer models designed to simulate how the proteins are likely to move. Their work is made possible by access to a supercomputer built by D.E. Shaw Research, a company founded by Shaw after he stepped back from his investment firm.
Known as Anton 2, D.E. Shaw's supercomputer is essential for Relay, allowing it to predict the behavior of hundreds of thousands of protein atoms over much longer slices of time than would be possible otherwise.
Computer models of protein motion led Relay to realize that FGFR1 moves differently than FGFR2, a finding the biotech hopes to exploit with one of its first drugs in development, RLY-4008.
Over a year and a half or so, Relay scientists predicted the folding of FGFR1 and FGFR2, studying whether would-be drugs fit into and stick to FGFR2 but not its twin. Their process was a hybrid of virtual screening — digitally testing potentially millions of compounds against the target — and laboratory work to confirm their model's predictions for the select few that seem most promising.
In this respect, Relay is not much different than other biotechs which use virtual compound libraries to sift for "hits" to targets they hope to drug. But by layering protein motion simulations into the process, the company hopes to do it more efficiently and, more ambitiously, discover new ways drugs could potentially work.
"All the jigsaw pieces are on the table, but no one has ever really put them together," claims Patel.
Testing in cells suggests Relay might be on the right track with RLY-4008, showing it to be 200 times more potent in blocking FGFR2 versus FGFR1. Pemazyre and Balversa, by contrast, were much less selective.
Relay aims to begin clinical study of RLY-4008 in patients with cancer later this year.
Two other drug candidates, both directed at well-known cancer targets, were also disclosed by Relay this year as the company prepared for its IPO. As with RLY-4008, information gleaned from protein motion models was a key component in their design, and is why Relay believes they might outperform other drugs in development.
A Phase 1 trial for one aimed at a protein called SHP2 began in January.

'The hardest scientific challenge'
In choosing cancer as the first proving ground for what it hopes to accomplish, Relay has set itself among tough competition. Cancer drug development has leaped forward in recent years, the result of advances that have little to do with protein motion.
Targeted drugs have already become standard treatments for tumors driven by mutations in a wide array of genes. More recently, biotechs like Array Biopharma and Loxo Oncology have made a name for themselves by designing therapies that are more potent and less toxic. And previously "undruggable" genes like KRAS have been put in play by advances in medicinal chemistry.
Relay's challenge is proving that mapping protein motion can take this fast-moving field farther than it's already come.
Understanding how proteins move seems like an obvious advantage. But it's not clear whether that knowledge is necessary in a world where other drug-hunting techniques are enabling the discovery and development of effective new therapies.
"Deciding which problems are enabled by this technology would be, I think, the hardest scientific challenge," said Boger, who wrestled with similar questions in building Vertex.
Take, for example, the development of antiviral drugs that block HIV protease, an enzyme critical for the virus' replication. The protein's structure was solved in the late 1980s, paving the way for some of the first medicines for HIV — an early success for structure-based drug design.
HIV protease, it turns out, has two flaps that swing around to cover the site where the drugs latch on to the protein. How those flaps move, and how their movement affects a would-be drug, are clearly important.
"You could do a magnificent scientific job of simulating the motion of HIV protease," said Boger. "But the problem was already solved without doing that."
Now that Relay has chosen its first three drug targets, the pressure is on the company to show that what its models reveal allow it to create medicines others can't.
"There are advancements happening everywhere," said Jami Rubin, a partner at the investment firm PJT Partners and a Relay board member, in an interview. "We have to pick our spots."
The next revolution?
In 1981, eight years before Boger founded Vertex, a Fortune magazine cover proclaimed that designing drugs with the aid of computers was part of the "next industrial revolution."
Over the four decades since, computer models have become enmeshed in drug discovery, with each new advance promising to solve the industry's high rate of failure. Then it was molecular modeling by computer. Now it's machine learning.

Drugmakers, looking for an edge, have integrated these computational tools alongside tried-and-true experimental approaches. Companies like Schrӧdinger, Insitro and Nimbus Therapeutics have built themselves around blending the two fields. By pitching protein motion as its calling card, Relay is doing so as well.
"Ten years forward, all small molecule drug development will be done with this combination of disciplines," said Relay's Patel.
But for all the potential promised by computational drug discovery, the industry hasn't yet solved its core problem. Most drugs that reach clinical testing still fail, regardless of which technology led to their creation or selection. Biology still bests software, at least most of the time.
"One of the stumbling blocks of technology-based companies is they sometimes get entranced with their technology because they don't appreciate the old days," said Boger. He suggested Relay, with its personal ties to Vertex and other well-tested drugmakers, is less likely to fall into that trap.
Even if Relay proves its method, though, early successes won't necessarily lead to an approved drug, or be easily replicated for others.
Boger would know. Vertex, for all its pioneering research, needed two decades before winning approval for its first drug. In between, the biotech experienced the types of setbacks that are the norm, not the exception, for the industry.
With protein motion, Relay may have a head start if it's chosen the right drug targets, Boger said. But it likely has a similar journey ahead of it. "The odds are still against you," he said.