Wondering what actually wins science fairs? This post looks at three real ISEF-winning projects from recent years: what they did and how you’d get started on something similar. Use them for inspiration, not to copy.
Science Fair Information
The International Science and Engineering Fair (ISEF) is the largest pre-college science competition, drawing thousands of top young scientists from over 80 countries. You present advanced research, compete for scholarships and grand awards, and meet other researchers and experts along the way.
Getting to ISEF is a long road that starts at local and regional fairs. That’s exactly why qualifying looks so strong on an application: it means you’re among the best in the world at your research. There are other science fairs worth entering too, but ISEF is the biggest and most competitive, so it’s the one worth understanding first.
Past Projects
Synthetic DNA Engineering With ICOR
Rishab Jain‘s project works in synthetic biology, improving protein production in E. coli, which matters for vaccine development. The core idea is codon optimization: choosing the DNA sequences that get a cell to produce the most protein. Standard methods often ignore how the cell actually behaves, which makes them inefficient. Jain built ICOR, a tool that uses a recurrent neural network (RNN) with a bidirectional long short-term memory (LSTM) architecture, trained on a dataset of high-expression E. coli genes, to optimize sequences in a way that fits the cellular environment better. Tested against standard methods, it measurably improved protein expression, with broad implications for biotechnology and vaccine development.
Materials and Requirements: Coding knowledge, a high-performance computing system that can handle large datasets, software for implementing the models, comprehensive genomic datasets, and statistical and data-analysis tools.
Getting Started: Build a solid foundation in bioinformatics and machine learning first. Then gather the computational tools and genomic data you need, and write a project plan with clear stages for training, testing, and validating your model. Refine the design as your results come in, and keep notes so you can document and present your findings.
Award: Regeneron Young Scientist Award (i.e. TOP 3, winning $50,000!) at ISEF 2022
BIO-PLEX: An Innovative Biocomputational Approach
In 2022, as the world was still dealing with COVID-19, Mpox emerged as a new threat with higher infectivity, leading to 30,000 cases in a short span. That raised urgent questions: what changed in the virus to make it spread faster, and would those changes affect how well treatments work? Saathvik Kannan built BIO-PLEX, a computational tool that studies the virus’s structure and mutations. It showed how two changes in the virus’s DNA-copying protein might make Mpox spread more easily and affect treatment, and it’s a clear example of how computational methods can study a disease quickly.
Materials and Requirements: High-performance computing resources, deep learning libraries, homology modeling tools, Python experience, virus genetic data, protein-structure and mutation databases, and software for mutation profiling.
Getting Started: Get a solid grounding in computational biology and the specific virus you’re studying. Learn deep learning and homology modeling, since those drive the structure prediction, and get comfortable in Python. Gather your genetic data and learn the relevant tools and databases, then start analyzing the virus’s structure and mutations to look for what drives its infectivity and any resistance to treatment.
Award: Regeneron Young Scientist Award ($50,000) at ISEF 2023
Self-Supervised 3D Human Motion Reconstruction
Michelle Hua‘s project reconstructs 3D human shape and motion from a single (monocular) video, getting around the problem that existing methods need large training datasets and still struggle with accuracy. Her method, a geometric consistency-based self-supervised neural network (GC-SSN), uses joints and silhouettes pulled from video frames. By forcing the reconstructed 3D models to stay consistent with those features, GC-SSN reaches high accuracy without manual labels or ground-truth data, and it outperforms current methods. It has promising uses in 3D broadcasting, virtual reality, sports analysis, and telepresence.
Materials and Requirements: A solid coding foundation, monocular video footage, a computer with enough processing power, image-processing software, a machine learning framework, training data (optional), and geometric-modeling libraries.
Getting Started: Get comfortable with the basics of image processing and machine learning. Learn how to extract and analyze human motion from video, and understand the principles behind geometric modeling. Read existing research in the area to see what’s been tried and where it falls short. From there, you’ll be ready to experiment with your own techniques.
Award: Regeneron Young Scientist Award ($50,000) at ISEF 2023
An Important Note
Don’t just copy these projects. The whole point of research is to make something original. Use them as a starting point for your own brainstorming, find a problem you actually care about, and build something new. That’s what wins.
Conclusion
Take a look at Rishab’s free STEM Student Guide, which has a lot of practical help for building a strong science fair project, plus other opportunities like competitions and research programs. You can also watch his YouTube playlist on how to do research.
Leave a Reply