Nystrom is bringing AI into biotech to create safer and more effective medicines.
Nicholas Nystrom, PhD, is a computational science trailblazer.
Driven at the age of 11 to educate himself on math, science, and computer engineering, he used college textbooks handed down from family members to learn calculus and college-level chemistry. He started his own training in computer engineering while in the 10th grade, learning to run certain software that required adding a different type of processor.
"By then I'd learned enough electrical engineering to make it feasible, and I ran that hybrid system for a couple years," he says. "I've never been happy with what computing has to offer. And so that's why I started designing computers to do things, so we can answer time-sensitive questions today, rather than waiting five or 10 years for the commodity market to give us what we need."
Nicholas Nystrom, chief technology officer of Peptilogics. Photo courtesy Peptilogics.
Nystrom received his PhD in computational chemistry from the University of Pittsburgh. Since then, he has created several innovative platforms, each enabling up to 30,000 users to conduct more than 2,500 projects, particularly in AI for the life sciences.
"I've been doing computational science for my whole career," he says. "I could see the ability of computational science to help us look at things we can't observe experimentally."
After 28 years leading scientific research teams at Pittsburgh Super Computing Center, Nystrom moved clinical-stage biotechnology company Peptilogics. He was excited by the opportunity to speed up lead compounds to patients.
"We're trying to get to those lead compounds as fast as possible," he says. "We are striving to make more of a difference."
He was also a part of the partnership between Carnegie Mellon University and the University of Pittsburgh, sponsored by the National Science Foundation (NSF), that created the supercomputer Bridges.
"Around 2014, I designed the first computer in the world that brought together high-performance computing, artificial intelligence and big data," he says.
This was at a time when high-performance computing had not yet been used by researchers in other fields, so Bridges was designed to enable these researchers to work easily with supercomputing.
Bridges beat the world’s best human poker players, improved predictions of severe weather to lengthen warning times, and offered gene researchers an easy-to-use tool to assemble the largest DNA and RNA sequences, according to Carnegie Mellon. In 2019, Bridges’ AI capabilities were enhanced with the latest GPU technology, fueling more sophisticated AI work on Bridges. In 2021, an advanced version, Bridges-2, was launched, integrating new technologies for converged, scalable HPC, machine learning, data, and more.
Nystrom was Peptilogics’ fifth employee, joining in 2021 as SVP and head of computation and data, with the goal to scale drug design using generative AI, HPC, and physics-based simulation. He was promoted to chief technology officer one year later.
He had met Peptilogics CEO and founder Jonathan Steckbeck in 2020 and discovered they had the same long-term vision: to use computational science to scale therapeutics design. Steckbeck's vision was to take what he had done through biochemistry and wet lab work to make that scalable through a machine learning approach.
"At Peptilogics, we recognize that AI is transforming the life sciences," Nystrom says. "Where we are focused today is in developing therapeutics. We are focused on being very general, being able to treat arbitrary targets and arbitrary therapeutic areas and that makes it scalable. That means we can go after much more in the long run than we were ever able to do historically. And that's what I was enthused to do."
Nystrom says the biotech is researching very diverse biological targets and diverse diseases that range from membrane proteins down through the target in the cell nucleus.
“We're looking at diseases covering rare disease or genetic disease, cancer and immunology, with others in the pipeline and we are focused on algorithms which led us to the capability to treat general targets and hence general therapeutic areas, rather than having a specific focus on the target class or disease,” he says. “In fact, we could not have done what we did just for one disease because there would not be the data algorithms that can work with finite biological data."
This progressive work environment requires a certain culture of open-minded thinkers, he says. As CTO, Nystrom has built these teams from the ground up.
"The team we have built is focused on people who are very inquisitive, who embrace continuous learning, because this field is moving so fast, and people who really want to make that transformative difference," he says. "It's a very interdisciplinary team.”
“As a leader, I bring a culture of thinking broadly, recognizing people have deep expertise in science and in machine learning, but that everyone is always learning something from others, because there's never anyone who's a master at all, including myself,” he adds. “The biggest challenge in this field is not the implementation, because we know how to do that. It's that continuous learning culture and finding the people that actually have this forward-looking mindset of doing things in a new, better way without saying this is the way I've always done it. And so that's what we hire for, people who embrace that constant curiosity."
This type of culture also requires tackling problems from a "monkey first" mentality, he says, referring to the theory of Astro Teller, the CEO of X, Google's innovation hub, who believes prioritizing the most difficult challenges of a project first is key to success. For example, if your objective is to have a monkey stand on a pedestal and recite Shakespeare, you start by teaching Shakespeare.
"If you build the pedestal first, you will feel like you're making progress because it's easy to build a pedestal," Nystrom says. "But in the end, the really hard thing is teaching a monkey to recite Shakespeare."
At Peptilogics, the principle is to start with the hard thing.
"We bring these different complementary pieces together between understanding science, understanding artificial intelligence, [and] understanding how to compute them, and make them run really well," he says. "And then get to work."
And there is plenty to work on. A 2018 paper, illuminating the druggable genome project, determined only 3% of known targets have been commercially drugged. It identified 62% of targets as having chemical or biological support, many of which Nystrom and his team expect to go after.
Development and application of machine learning architectures and models will create safer, more efficacious medicines and help us to understand key aspects of systems biology that drive disease.
"That's where AI-driven design can potentially make a truly meaningful difference," Nystrom says.
“There's just so much to learn, I can't help myself. Learning is just perpetual for me.”
— Nick Nystrom, CTO, Peptilogics.
Robin Robinson is a contributing writer for HealthLeaders.
A self-taught computer engineer, Nick Nystrom tackles some of drug design's biggest challenges with computational science.
He joined Peptilogics after 28 years in academia, seeking to change how drugs are designed.
The biggest challenge in using AI, he says, is creating a team that can keep up with the innovative thinking.