Enabling AI-Powered Computational Biology in Pursuit of Precision Medicines

Debojyoti “Debo” Dutta, vice president of engineering, AI at Nutanix leads artificial intelligence efforts to accelerate the development of new therapies and enterprise productivity.

By Tom Mangan

By Tom Mangan March 14, 2024

Where the rest of us see medical mysteries, Debojyoti “Debo” Dutta sees irresistible math problems. He’s devoting his life to using computer science to solve them.

Dutta is vice president of engineering (AI) with Nutanix, a hybrid multicloud IT infrastructure software company that helps enterprises modernize and scale their digital capabilities. He also has a fascinating passion that converges mathematics, computer science and biology to drive a new generation of personalized medicine powered by artificial intelligence (AI).

In an interview with The Forecast by Nutanix, Dutta talked about his decades-long passion for computational biology, which uses statistical algorithms to unravel riddles buried deep in human DNA and proteins. 

Dutta played a pivotal role in developing an open-source software platform to strengthen medical research by improving the accuracy of models that enable artificial intelligence and machine learning (AI/ML). In the interview, he described where AI/ML is poised to drive significant gains in therapeutics, and explained why digital infrastructure technologies will be central to securing those gains.

Where It All Started

Like many technologists of his generation, Dutta started studying computer science at the India Institute of Technology. But while his professional work would dwell primarily on ones and zeroes, he nurtured a fascination with living things. 

“I kept thinking about biology, but I didn't do much about it,” he recalled. 


Quest to Improve Cancer Therapeutics with AI and Computer Science

That changed when he entered Ph.D. studies at the University of Southern California (USC), where the mathematician Mike Waterman cofounded the field of computational biology.

“That's when my eyes started opening up,” Dutta said. “And after my Ph.D., I said, you know what? No matter what I do in life, I need to spend some time in computational biology.” 

So, he did three years of post-doctoral research that helped him learn to develop the elaborate statistical algorithms that produce what we now call machine learning.

“I got to learn about machine learning as a part of understanding how the human body works,” he recalled. When he left USC and went into private industry, it wasn’t long before he applied his knowledge of statistical algorithms to improve networks at Cisco Systems.


Safe Place to Stress Test Medical AI and ML Apps

While his day job had him polishing his networking and cloud computing skills, after work, he kept returning to a theme from post-doc days: Using computer science to improve cancer therapeutics. So, he made up his mind to change things.

In 2018, Dutta joined a collective of AI/ML experts as a founding member of what became MLCommons, which was established to create performance benchmarks, datasets and best practices for machine learning. Within MLCommons, Dutta co-incubated MedPerf, an open-source platform that helps medical researchers confirm the reliability of AI models without endangering sensitive personal data.

MedPerf uses a federated learning framework that allows researchers at a hospital or university laboratory to assess the models within a medical institution without having to give out their private data, which is also known as federated learning systems. 

“No data leaves the premises of the healthcare organization — it stays within their private network,” said Alexandros Karargyris, a machine learning researcher who helped spearhead the MedPerf project, in an interview with The Forecast. Karargyris was lead author and Dutta was co-author of a research report in Nature Machine Intelligence explaining how MedPerf works.

Data, DNA and the Future of Medical Discovery

The MedPerf platform addresses two questions that torment users of AI/ML in medical research: How can they be sure AI/ML models produce statistically valid results, and how do the models protect sensitive personal data against hackers and cybercriminals?

MedPerf does this by establishing benchmarks for reliable, consistent AI modeling and creating a platform that essentially delivers the models to researchers. Patient data remains in-house, where researchers can keep it safe and secure.

“Models are the heart of machine learning,” Dutta said. 

“You use data to train models, and then you deploy the models to infer from new data.” Data quantities grow bigger every day, driving the demand for high-quality, unbiased models. Models also must comply with data protection rules and follow established best practices. MedPerf helps make all this possible.


IT Leaders Get AI-Ready and Go

Dutta explained that AI is advancing medical knowledge in three principal ways — imaging, DNA and protein sequence analysis and generative AI-based rapid experimental research. AI/ML algorithms can scan X-rays and other images for signs of tumors and other anomalies. Retinal scans, for instance, track the progression of diabetic retinopathy.

A large language model (LLM) like ChatGPT could scan massive libraries of medical journal articles and published research studies to help doctors diagnose ailments more quickly. LLMs also can comb through terabytes of DNA-sequencing data to help uncover the genetic triggers of diseases.

Dutta noted that interdisciplinary researchers are exploring using LLMs along with CRISPR and immunotherapy to destroy cancer cells. (Editor’s note: Dutta shared links to related articles published by Serafim Batzoglou, The National Library of Library of Medicine and Nature Portfolio.)

“Using AI, you can rapidly design or search DNA sequences that matter for cancer therapy,” Dutta said. “What would've taken us years now might take us a tenth or a hundredth of that time.”  


Generative AI Speeds Protein Design

It doesn’t take a clinician or researcher to see the value AI brings to medicine. 

“As a computer science systems person, if I can help accelerate biologists to go faster with better AI-based tools, then these experts can make cancer better and better — and we can cure more types of cancers and other complex diseases,” Dutta said.

The Role of Digital Infrastructure in Enterprise AI

Nutanix started by pioneering hyperconverged infrastructure (HCI), which uses software to emulate the hardware functions of compute, network and storage. Dutta arrived at the company in 2020 in the depths of the COVID-19 pandemic and quickly saw how HCI offered the flexibility and scale needed for AI/ML computing environments.

“Nutanix can host a lot of amazing next-gen AI workloads that can change the world in ways that we don't even know,” Dutta said. 


Building Infrastructure to Unlock AI’s Next Big Leap

The arrival of LLMs like ChatGPT encouraged the company to develop a purpose-built solution after enterprise customers started asking for AI tools they can run on-prem without having to share data with an outside firm.

“Now, our customers can run generative-AI models, including LLMs that fuel applications like ChatGPT, on their private infrastructure, with their own data and on their own terms, with the right safety and governance,” Dutta added.

Proper enterprise infrastructure frees large companies to use LLMs to scan legal documents and current legislation to ensure compliance and reduce liabilities. LLMs also can streamline customer service with chatbots and other conversational interfaces. And these capabilities barely scratch the surface of what’s possible.


Exploring Uses for AI in Healthcare

“I think this is going to dramatically change how the future enterprise will operate,” Dutta said. “And that's why we are still underestimating its economic impact.”

Dutta’s contributions to and championship of MedPerf and MLCommons enable innovations across the field of medicine, opening potential for new drugs and using IT infrastructure to create digital twins of entire treatment-development processes. It’s all part of a once-in-a-lifetime opportunity to use advanced automation to enhance (rather than replace) the innate strengths of our species.

He’s optimistic about what’s in store: “I believe that AI might actually make human beings more intelligent and efficient rather than the other way around.”

Editor’s note: Learn more about Nutanix GPT-in-a-Box, a full-stack software-defined AI-ready platform designed to simplify and jump-start your initiatives from edge to core. More details in this blog post The AI-Ready Stack: Nutanix Simplifies Your AI Innovation Learning Curve and in the Nutanix Bible.

Tom Mangan is a contributing writer. He is a veteran B2B technology writer and editor, specializing in cloud computing and digital transformation. Contact him on his website or LinkedIn.

Jason Lopez and Ken Kaplan contributed to this story. Lopez is executive producer of Tech Barometer, the podcast outlet for The Forecast. Ken is Editor in Chief for The Forecast by Nutanix.

© 2024 Nutanix, Inc. All rights reserved. For additional information and important legal disclaimers, please go here.