Alumnus Develops ‘ChatGPT for Genetics’

What began as a bold research idea at Emory is now Bystro AI, Alex Kotlar’s effort to make genetic data easier to understand and use.

When Alex Kotlar 18PhD started graduate school at Emory 13 years ago, he envisioned a Google-like search engine for genetics. He imagined being able to ask a computer questions in plain language, such as:

How does this person’s genome differ from their parents?
Are any of those variants known to cause cancer?
Could any of these changes affect how this patient responds to medication?

Kotlar’s vision was ahead of its time, pushing beyond what existing tools and infrastructure could readily support. But now, Kotlar has launched Bystro AI, an artificial intelligence research assistant that’s based in Boston at the MassChallenge accelerator program, which has empowered more than 4,500 startups since 2010.

Bystro AI was funded by a donation from Gates Ventures through Emory, and the company already has contracts with Emory, UC Davis and an inflammatory bowel disease consortium.

The pitch Kotlar makes to fellow researchers is: Bystro lets you focus on scientific questions by handling the implementation details. Programs like it could already take the raw data from a genome or thousands of them and manage all the genetic variants and annotations, but the latest version of Bystro can dramatically simplify complex tasks that would usually require someone with extensive programming experience.

“It connects together all of the tools you would want to use to analyze omics data,” he says. “For researchers, it’s like the highly trained postdoc you’d want to hire, ready to work for you.”

Beyond specialized research, Kotlar thinks Bystro could become a tool for physicians or consumers at home to analyze individual genomes and understand their risk for diseases or which medications could be best. At a time when many people have had their DNA analyzed through services such as Ancestry.com, it’s a vision that seems within reach. Early users have reported gaining clearer insights into how genetic variants may relate to areas such as diet, medication response or underlying health considerations, often as a starting point for conversations with clinicians.

“We want to get it into the hands of more people,” Kotlar says. “I think the technology is finally living up to the promise.”

SEEDS FOR GROWTH PLANTED AT EMORY

Kotlar came to the United States when he was five years old, an immigrant from Ukraine whose family was fleeing the Chernobyl nuclear power plant disaster. Several of his family members died from cancer, and Kotlar was raised being told he had a 50% increased risk of developing cancer during his lifetime. That was partly what drove his interest in genetics.

“Emory was a place where I could have a weird idea and people let me run with it,” he says.

Kotlar credits his graduate school mentor David Cutler, professor of human genetics research in Emory’s School of Medicine, with giving him space to work on his ideas and also with giving him critical advice about how to structure and store data.

Cutler recalls being uncertain at first whether adding a search capability was the most practical direction for the project. “[Alex] ignored me and produced the natural language processing algorithms that Bystro uses for search,” Cutler says. “It turned out to be a good decision.”

Kotlar, fellow graduate student Cristina Trevino 20PhD and several Emory faculty members including Cutler, published an early version of Bystro in 2018. Another co-author was Thomas Wingo 04M, an Alzheimer’s disease researcher who specializes in making sense of large data sets from consortiums that collect hundreds of brain tissue samples. Wingo, who moved from Emory to UC Davis in 2024, continues to advise Bystro as the company’s chief scientific officer.

After finishing his doctorate, Kotlar then moved to the Broad Institute in Massachusetts and then to the private sector. This helped him make business connections and led to launching the company, and he recruited Trevino to join him.

GETTING THE INFO INTO MORE HANDS

Kotlar sometimes describes Bystro as “ChatGPT for genetics,” but then acknowledges that the analogy is imperfect. In the genetics research community, Kotlar has encountered skepticism about large language models’ tendency to generate plausible-sounding but inaccurate results, sometimes called hallucinations.

He emphasizes that Bystro is not simply a text-based large language model, but rather a system that coordinates specialized analytic tools designed specifically for genomic data.

A staggering fact: Large-scale genomic studies now routinely involve billions of data points — volumes of information that dwarf even the largest digital libraries.

Kotlar also has to handle concerns about privacy and data security, amid growing public scrutiny around how genetic information is stored and used. Bystro doesn’t retain the genetic data that fuels the platform’s findings, he says. Instead, data are processed in secure environments, transformed into the formats needed for analysis and then removed, with results intended to support — not replace — clinical decision-making. In addition, every response generated by Bystro comes with a disclaimer: the findings should be reviewed with a physician or pharmacist. They should not be used to self-diagnose or self-medicate.

For Kotlar, the work is both technical and deeply personal. The child who once grew up hearing about inherited cancer risk is now building tools designed to make genetic information clearer and more actionable for others. What began as an ambitious idea at Emory has grown into a platform with national reach — one grounded in the belief that better questions can lead to better answers.