Insulin for diabetes. Lactase for gastrointestinal tract disorders. Factor VIII and IX for hemophilia. There is no doubt about the therapeutic potential of human proteins.
With this in mind, Flagship Pioneering’s Origination Partner Lovisa Afzelius, PhD, Vice President and Science Partner Hozefa Bandukwala, PhD, and Senior Principal Theonie Anastassiadis, PhD, asked: Are there potential sources of novel and powerful therapeutics yet to be revealed in nonhuman organisms?
Anastassiadis picks up the story: “Immediately, viruses came to mind because they’re rapidly evolving to be able to drive and modulate complex processes. In doing so, they actually confer on their viral proteins unique properties or characteristics. For example, it can be enhanced potency toward a certain target, cell specificity to target one over another and stability, etc., so that they can most optimally modulate that biology. Those are really the same features that we try to optimize when we’re developing biologics. So, then, we asked ourselves, Why not leverage these evolutionarily optimized viral proteins and repurpose them to develop therapeutics to treat diseases?”
Moreover, the viral proteome boasts significantly more proteins than the human proteome. Whereas humans have about 20,000 known proteins, the viral proteome—the complete set of proteins that all viruses make—contains around 6.2 million and is constantly expanding while being continually sequenced. That is a rich and vast catalog of proteins that may hold the key to some innovative therapeutics.
Based on this idea, Afzelius, Bandukwala, and Anastassiadis launched Prologue Medicines to focus on rapidly creating powerful new medicines by unlocking the therapeutic potential of the viral proteome.
The company created the DELVE (Decoding Evolutionary Logic of Variant Ensembles) Platform to use tools from artificial intelligence (AI) and machine learning (ML) to systematically investigate new proteins and unique features of the vial proteome.
Flagship has initially committed $50 million to support Prologue, which aims to develop a pipeline of medicines for various diseases.
Multimodal knowledge
To find therapeutic molecules in the viral proteome, Prologue can’t just apply the latest and greatest in AI and ML and hope that it produces hits—it requires context.
According to Bandukwala, who is the Chief Scientific Office at Prologue, researchers have been sequencing organisms on a massive scale and their entire proteome is now available. But all of these sequences get deposited in databases where the original researcher probably only has an interest in a handful of those—the rest of the sequences stay there. To make these sequences useful, Prologue has built an underlying multimodal network that brings in a variety of information—essentially, data on human proteins, existing proteins, characterized proteins, drugs that target these proteins, diseases associated with these proteins, and genome-wide associations involving these proteins.
“Other people have been looking at viral proteins as well, but essentially, that has been anecdotal or one protein at a time… I am interested in a particular virus and a particular protein and will characterize this,” Bandukwala said. “We go way beyond a simple homology search and think about various attributes that we can utilize to start thinking about these proteins. We can mine the entire viral protein systematically.”
This information network has culminated in a computational tool known as a knowledge graph. This enables the exploration and analysis of multimodal data that is impossible with the human eye.
At the end of this complex in silico process, an experimental process begins to validate the output of the knowledge graph. Anastassiadis said that Prologue has quickly brought in the latest innovations and automation to establish structure-activity relationships.
“We can now do almost any kind of functional assay at scale,” said Bandukwala. We can produce these proteins at scale and characterize them at scale.”
Losing my religion
According to Bandukwala, the beauty of it all is the DELVE platform. At the same time, it can be used to look for candidates with human targets and can also harness the creativity of viruses.
“We sometimes get very linear in our thought process: I need to achieve goal A, so I must do B,” said Bandukwala. “But looking through viruses, we can really see how innovative evolution has been and doing things in an extremely unique manner.”
Bandukwala provides an example from the literature that shows that tumor necrosis factor (TNF) is an excellent target for autoimmunity. It turns out an entire lineage of a poxvirus family has been utilizing inhibition of TNF as a mechanism for immune escape, creating a variety of different proteins to block the action of TNF.
“These viruses evolved millennia ago… and they have figured out how to inhibit TNF by binding to the exact same site where Humira binds,” said Bandukwala. “[The viruses] figured out way earlier than we did the best way on how to inhibit TNF.”
Bandukwala finds it remarkable that the viruses don’t use an antibody format but a completely different protein. That’s exactly the kind of viral creativity Prologue is trying to tap into.
“We are finding that these proteins have been refined further, and essentially this entire process that we know happens in biologics development—that you start with something and then you have to optimize it—and further, we are finding that evolution has already taken care of that,” said Bandukwala. “Once we find a really good target, we anticipate that we will actually have a pretty rapid timeline to take it into development.”
With DELVE, Prologue takes an agnostic approach to decoding the logic of viral protein variant groups, or “variant ensembles,” to find proteins that perform exceptionally well or have completely novel features not seen in human proteins.
“As we look across this entire viral proteome, we find that it’s rarely the case that when there is an important piece of biology, you find multiple different viral proteins that seem to have similar biology,” said Bandukwala. “That is one of the biggest powers we can leverage that we are looking at; it is across not a single protein but the entire variant space.”
With the high-throughput experimental platform, Prologue can examine all the variations in different proteins, map them out to function, and get detailed resolution on the key attributes in the various proteins that are driving those different functions.
“This becomes a critical part of how we think about developing them into therapeutics,” said Bandukwala. “It’s not just that this protein does this better, but it’s understanding why it does that better. That helps us map that entire development pipeline.”
Nonvirtual reality
By understanding all of these viral ensembles, Prologue can go leaps and bounds beyond looking for unique bits of viruses to make preventative medicines like vaccines and instead look at how they can be used therapeutically.
“Historically, people have studied this area with an eye towards prevention of diseases,” said Anastassiadis. “This is taking that and flipping it on its head and asking: If these viruses have optimized the best way to modulate human physiology, why not leverage that itself and turn it into medicines that could highly impact patients?”
“We think viral proteins have a lot of power because they’ve evolved in ways that human proteins haven’t, and they’re constantly under these selective pressures to optimize and engage with that host biology.”
Anastassiadis said this isn’t just a theoretical thought experiment—Prologue has already experimentally tested dozens of these in silico-predicted viral proteins.
“What is really in our favor is that these proteins are not hypothetical,” said Bandukwala. “As we know, some newer [AI] technologies can just hallucinate proteins. There are also many hypotheticals because you have no idea what they’ll produce. These [viral proteins] already exist in nature. They are functional components of organisms that are there in nature… Our success rate turns out to be pretty high. In terms of how proteins are produced, they have activity in most cases.”
According to Anastassiadis, viral proteins that have undergone evolutionary optimization inspired the company’s name.
“Evolution is really the basis on which the prologue is really leaning,” said Anastassiadis. “So, it is the prologue to this upcoming chapter on using viral proteins as therapeutics.”
