Welcome to Trailblazers, a series featuring pharma leaders who’ve carved new scientific and business paths for the industry. Today, we’re featuring Jordana Blackman, co-founder and CEO of Australia’s Omnigeniq, a computational biology company with a protein-folding software that stands apart from others in the field.
Computational biology promises a deeper understanding of protein structure and function in disease and medicine. A Nobel Prize in 2024 for Google DeepMind’s protein-folding platform AlphaFold planted a flag in the evolving landscape — but where does it go from there?
Models like AlphaFold and others that depend on massive datasets in the known universe of human proteins open new doors for drug discovery and development, said Jordana Blackman, CEO and co-founder of Australian computational biology company Omnigeniq. AI giants like Nvidia, Microsoft and Google all have their hats in the ring, and there’s room for innovation.
“Computational biology evolved along with the digital age alongside lab work and amazing progress at research institutions around the world,” Blackman said. “Right now AI [large language models] are excellent tools that open up hundreds of thousands of new shapes and models that scientists are able to work with, but we approach it a little differently.”
Blackman and fellow co-founder Tiffanwy Klippel-Cooper met on a space technology project designing a bioreactor to grow pancreatic organoids that would produce insulin for astronauts. In the confined conditions of deep space travel, these organoids needed to be self-sufficient by recreating in vivo conditions with as much fidelity as possible, Blackman said. While solving for these constraints, Blackman and Klippel-Cooper realized just how good their software had become at creating proteins in a self-contained system.
"We’re connecting dots in an entirely new way.”
Jordana Blackman
CEO, co-founder, Omnigeniq
In the interest of pushing their technology further, they began uploading drug files to see what would happen.
“With pure scientific curiosity, that was the beginning of our new and current direction of in silico computational models,” Blackman said. “Ours is the first model that recreates the conditions in which a protein emerges. As opposed to crunching numbers in the traditional way, we’ve created the conditions in which cells create a protein in their native state.”
Combining biophysics and AI through a concept called “deterministic intelligence,” Omnigeniq’s platform overcomes the data constraints of probabilistic AI models. The platform doesn’t rely entirely on external training data, but rather an internal framework and architecture to mimic the living conditions in which proteins form.
“We’re able to visualize for the first time what these proteins look like in their ligand-ready state, before they’ve been conjugated or bound to any molecular supports,” Blackman said. “This is a huge step forward, and we hope to be able to use it as a tool to design or engineer therapies that come from an understanding of the changes that lead to disease propagation.”
Omnigeniq is starting small — microscopic, to be precise — as they aim to be the first company to build a computer model of an E. coli organism. The concept would “prove the power of computational modeling” and bring the company closer to its long-term mission of creating the world’s first holographic twin of the human body.
In the meantime, Omnigeniq is in discussions with what Blackman described as a “leading research institute” to evaluate toxicology and compatibility of four assets — an effort that could help usher the candidates into human trials. They are also working with an undisclosed U.S. neurodegenerative disease nonprofit to help understand the illness’ mechanism of action.
“It may be that we’re uniquely placed to be able to show how, where and why that specific disease propagates,” Blackman said.
Broadening druggability
Druggability is defined by target receptors on proteins that can be modified using a uniquely shaped drug, and researchers have gotten very good at finding them. But out of 20,000 human proteins, only about 700 of these receptors are accounted for, leaving a wide range of possibilities for new approaches and a changing definition of the concept of druggability, Blackman said.
The way proteins are traditionally seen, through X-ray crystallography or electron microscopy for example, provide a map of the molecule in a certain state, and perhaps not its natural one, Blackman said. Being able to see more of these receptors via a model that retains their functional shape could point researchers in new directions.
“Imagine broadening that definition by even just one order of magnitude and what that would do for cutting-edge, innovative small biotechs who are able to go after new, exciting targets, and how that would flow through the entire pipeline,” Blackman said.
But breaking through to the biopharma industry, which is notoriously slow at adopting new technology, is another matter.
“The pharmaceutical R&D industry must remain conservative because it’s no small thing to develop life-saving therapies for people with diseases,” Blackman said. “So while technology developments are exciting and scientists are eager to test and adopt them, it needs to be at a pace that’s right for regulators and for R&D teams to use them the right way.”
While full adoption of computational models will take time, some aspects of drug development could more readily make use of them, especially for smaller biotech companies without multimillion-dollar compute budgets, Blackman said.
From data to causality
Data will always be king in drug R&D as the basis of the scientific method. But while evidence of causality is traditionally built through experimentation and repeatability, the Omnigeniq model applies causality in reverse to add a layer of context, Blackman said.
“Without generating millions of data points that costs millions of dollars per experiment, to be able to show through a computer model that X leads to Y leads to Z, we are making sense of data we haven’t seen yet,” Blackman said. “There may be points of causality that can’t be known to a pattern-matching model — we’re connecting dots in an entirely new way.”
Omnigeniq isn’t trying to make data obsolete in drug R&D. The company is making each data point more useful and efficient by extrapolating the causality of protein folding in an in vivo computer model, Blackman said.
Big steps in technology have a tendency to generate a great deal of optimism and hype. The completion of the Human Genome Project in 2003, for example, came with high hopes for immediate solutions to long-held problems. While the information was a monumental undertaking, its greater impact was providing tools for scientists to explore further, Blackman said.
Computational biology has entered the scene in a similar way, she said. With LLMs in general and models like Omnigeniq’s, the answers don’t come neatly wrapped for immediate gratification. Scientists and researchers will continue to push their work further, only now with more tools at their disposal.
“Alongside the commitment to observe and feed back into a model and have it learn from real data, our model can show how molecules and proteins behave in vivo as opposed to proving it in a plate,” Blackman said. “Disrupting the huge attrition rate in clinical trials is where we hope our model will be able to assist.”
The lesson to be learned from technology disruption in biopharma is that “no machine will ever be able to do what the human mind does … but by giving scientists better tools so they can do more,” Blackman said.
As Blackman and the Omnigeniq team speak to drugmakers about their model, a lot of time is spent educating about the new approach.
“Where all the trends are going in this big data and hyper-scaled age, our approach is so different,” Blackman said. “But we’ve had an immense amount of interest in our compute method because it’s deterministic in nature, not probabilistic like AlphaFold’s, and it requires a lot less computational space.”
With demonstrations possible on a regular laptop, Blackman said Omnigeniq hopes to alleviate the “crushing weight” it takes to be a part of the LLM and AI-based computational biology revolution.
