Melissa Pappas
When he enrolled at Dartmouth, Pranam Chatterjee, Assistant Professor and Africk-Lesley Distinguished Scholar of Innovation in Engineering with a joint appointment in Bioengineering (BE) and in Computer and Information Science (CIS), did not start off as an engineering student. He majored in religion.
“I had a deep desire to understand the people around me,” says Chatterjee, who grew up in a household of scientists on the Alabama-Georgia border.
But after transferring to MIT, where he returned to his science roots and studied biology and computer science, Chatterjee brought the unique perspective of the humanities with him.
“Studying religion taught me to be a people-first engineer,” he says. “No matter how powerful your technology is, if it doesn’t fit into the way people see the world, it won’t have the impact you want.”
From the Humanities to AI for Biotech
Chatterjee quickly turned to artificial intelligence (AI) as his tool of choice in the engineering world, asking himself the question,”How can AI be used first and foremost to solve the most pressing problems facing humanity?” The pursuit of the answer to this question cleared a path that would lead him to the forefront of AI for biotechnology.
“Biotechnology has the potential to solve so many of the world’s most pressing problems, from health and medicine to energy, agriculture, food security and the environment,” Chatterjee says. “When designed well, AI can make sense of biological complexity and design solutions at scale.”
Today, Chatterjee is a member of the Center for Precision Engineering for Health, where his lab develops generative AI models to identify and design molecular therapeutics that have the potential to treat diseases considered “undruggable,” like pediatric cancers, which are caused by unstable, malfunctioning proteins. These diseases are hard to target and synthesize, making the endeavor challenging.
But Chatterjee saw the problem and turned it into an opportunity that AI could help solve.
Generative AI for Biology’s Edge Cases
In his Penn Engineering lab, the workflow is strikingly direct. Publicly available biological data — often from patient samples or past clinical trials — feeds into advanced generative algorithms. These models don’t just predict properties of molecules; they create new ones from scratch, designed to bind to and disrupt mutated or biologically malfunctioning proteins which can cause diseases such as neurodegeneration, viral diseases and many forms of pediatric cancer. They even have recently extended their models to targeting toxic heavy metals.
Unlike structural approaches to creating disease-treating drugs, which rely on 3D models of proteins, Chatterjee’s team works purely from sequence data. Their algorithms generate amino acid chains in both directions, akin to a language model that can compose a sentence by filling in words before and after a given point.
On top of identifying patterns in malfunctioning proteins and designing molecules to treat those problem areas, the algorithms that Chatterjee leverages add another layer: multi-objective optimization, ensuring that a candidate molecule isn’t just effective against its target, but also non-toxic, stable and orally bioavailable. Multi-objective optimization here is akin to what we do every day when weighing all the factors that go into making decisions that need to meet multiple goals — it’s a trade-off to find the best solution that works in the real world.
“Biology needs multi-objective optimization,” Chatterjee explains. “For a therapeutic to work in the clinic, it has to meet many properties at once. We build that into the generative process from the start.”
Tackling the Rare and the Difficult
The advantage of doing this work in an academic setting rather than in a clinic or a startup drug development company is that Chatterjee’s group can focus on diseases that industry often avoids. These include conditions too rare to be profitable — such as Alexander disease, a genetic disorder that affects the brain and spinal cord — or too complex to treat with conventional drugs. For example, pediatric sarcomas caused by unstable fusion oncoproteins, or “Frankenstein” proteins that get created when two different genes accidentally fuse together, are some of the focus areas in Chatterjee’s lab that other industries would not take on.
“I’ve learned that companies mostly focus on problems with clear commercial markets, but academia gives us the freedom to take on the cases that others overlook,” Chatterjee says. “Rare diseases driven by hard-to-drug proteins may not be profitable to pursue, but patients depend on us to solve them. This is the perfect setting to integrate new AI into biotechnology and deliver therapies that patients would otherwise never receive.”
Huntington’s disease and even substance use disorder are also target conditions for Chatterjee’s algorithmic approach to drug discovery. In each case, his lab’s models generate therapeutic peptides designed for these stubborn targets, which experimentalists in the group then test in cells and animal models.
The Chatterjee Lab is a true hybrid — half AI researchers, half experimental biologists — working as a single team. AI specialists learn enough biology to see the therapeutic significance of their algorithms; experimentalists grasp the computational principles driving their projects.
“It requires everyone to get out of their comfort zones,” Chatterjee says, “but it’s the only way to make this work.”
From Bench to Bedside Without Leaving Campus
Penn, he continues, offers a rare environment where an idea can travel from theoretical math to preclinical testing without leaving the University. That’s critical for someone who has already spun out multiple companies like Gameto, a fertility startup with technology that has already contributed to births in multiple countries.
“Penn is one the best places in the world to take theories to real-world use,” says Chatterjee. “Here, I can take technology all the way from the lab to patients without having to build a startup every time,” he says.
In the years ahead, Chatterjee hopes his lab’s generative models will help rewrite the boundaries of drug design. But for him, success will always be measured not just in algorithms published or companies launched, but in people helped.
“If my technology can be positively beneficial to anyone, anywhere, that’s the best outcome.”
Applying Lessons From the Lab to the Classroom
Those technologies and Chatterjee’s people-first engineering approach are not just helping patients, they are shaping the next generation of engineers at Penn. Chatterjee will be teaching the undergraduate courses BE 3060 (Cell Engineering) and a CIS course on Discrete Generative Models over the next two semesters.
“I am truly glad to be at Penn Engineering because it gives me the opportunity to introduce these critical areas to young students and hopefully inspire them to do this kind of work in the AI and biotech space,” he says. “We need talented and interdisciplinary minds to tackle the biggest issues we are facing today, and just as Penn attracts driven researchers from around the country and the world, it’s also a haven and launching pad for some of the brightest students and future problem solvers.”
Learn more about the innovative work coming out of the Chatterjee Lab on his website.