🇺🇸 Neel Moudgal

The problem he chose

An RNA molecule's sequence — the order of its bases — is straightforward to determine. What that string folds into is not, and the folded shape is what determines function: which molecules it can bind, which reactions it can catalyse, what a drug might do to it.

Moudgal's entry was a computer model that predicts RNA structure using accessible data, aimed at disease diagnosis and treatment. The phrase "accessible data" carries the weight. Structure prediction methods that require expensive experimental input are useful to well-funded laboratories; one that works from data most researchers already have is useful to everyone else.

First place in the 82nd year

The Regeneron Science Talent Search was founded in 1942 as a programme of Society for Science, which makes it the oldest science and mathematics competition for American high-school students. In 2023 nearly 2,000 students entered, 40 were named finalists, and the finalists shared $1.8 million; the total distributed, including school awards, was $3.1 million.

Moudgal's $250,000 was the largest individual award. The gala was livestreamed and hosted by the broadcaster Soledad O'Brien. Regeneron, which took over sponsorship in 2017, had committed $100 million over ten years to the competition.

The competition's format is what distinguishes it from a science fair. Entrants submit an original research report, and the 40 finalists spend a week in Washington being questioned by a judging panel on scientific reasoning well outside the field of their own project.

The rest of the top ten

Second place and $175,000 went to Emily Ocasio of Fairfax, Virginia, who used artificial intelligence to measure racial disparities in how the Boston Globe wrote about homicide victims between 1976 and 1984. Third and $150,000 went to Ellen Xu of San Diego for an algorithm that uses smartphone photographs to help diagnose Kawasaki disease.

Below them came work on q-calculus, thyroid hormone treatment for brain injury, injectable microbubbles for stroke, machine-learning theory, immune reactions to head trauma, suicide-risk assessment from journal entries, and black-hole oscillations. Maya Ajmera, the Society's president and chief executive, said of the finalists that she was "in awe of their creativity and conviction."

Why RNA structure is worth the effort

RNA has moved from a supporting role in molecular biology to a central one. It is the basis of a class of vaccines, a target for drugs, and a tool for editing genomes, and every one of those uses depends on knowing what a given sequence will fold into.

Protein structure prediction has been transformed in the past decade by computational methods; RNA has proved harder, because the same sequence can adopt different shapes depending on conditions and because the experimental data available for training is far thinner. A model that works from accessible inputs is therefore attacking the bottleneck rather than the showcase.

There is a practical dimension as well. Designing an RNA therapeutic means knowing whether a sequence will fold into the shape that works inside a cell, and today much of that is settled experimentally, one candidate at a time. A reliable predictive model compresses that loop, which is why structure prediction attracts the effort it does.

What is documented — and what is not

Firm: first place, the $250,000, his age, his home town and the description of the model, published by the Society for Science in March 2023.

Open: the model's accuracy, its name, whether the code was released, whether a paper followed, and what Moudgal has done since. No source gathered for this page answers those questions, and none is answered by inference here.