Inspiration
Fungi are microscopic powerhouses used in biotechnology to produce everything from life-saving medicines to sustainable food ingredients. However, figuring out the perfect "diet" (or nutrient environment) to make a specific fungus thrive and produce what we need is incredibly complex. Instead of researchers spending months running physical trial-and-error experiments in a lab, we were inspired to create a system that allows them to test different nutritional recipes quickly and safely on a computer.
What it does
MediaOpt is a user-friendly web platform that simulates how fungi grow and metabolise nutrients. Users simply upload the basic biological data of a fungus, and the platform automatically builds a "digital twin" of its entire metabolism. Through a simple, slider-based web interface, researchers can adjust the virtual food supply and instantly see a prediction of how the fungus will react. It also lays the groundwork for an automated discovery tool that rapidly cycles through virtual environments to hunt down the absolute best conditions for that specific organism.
How we built it
We split the project into a modern web frontend and a heavy-lifting scientific backend:
• Frontend: We built a smooth, single-page web app using Vite and Vue 3, allowing users to upload files and tweak growth media via simple sliders.
• Backend: We used Python with FastAPI to handle the data. For the biological simulation, we integrated powerful bioinformatics tools: CarveMe automatically constructs the metabolic models from the uploaded data, and COBRApy runs the actual growth simulations.
• Optimisation Sandbox: We utilised a tool called SMAC3 to handle the smart, automated search over different media boundaries to find the best nutrient combinations.
Challenges we ran into
One of our biggest hurdles was bridging the gap between highly technical, command-line bioinformatics tools (like CarveMe) and a fast, accessible web interface. Translating complex biological constraints into simple software logic required careful orchestration. Additionally, designing a real-time optimisation engine that searches through thousands of diet possibilities without slowing down the web app proved challenging, which is why we focused on a placeholder system for the UI while building the search logic offline.
Accomplishments that we're proud of
We are incredibly proud of successfully automating the entire pipeline. We took a process that normally requires deep technical expertise going from a raw biological sequence to a fully interactive metabolic model, and made it accessible through a clean, visual web interface. Building a working prototype that successfully runs virtual biological experiments during the hackathon timeline is a massive win for our team.
What we learned
We learned a tremendous amount about integrating specialised scientific Python libraries with modern web frameworks. We also gained deep insights into the intricacies of genome-scale metabolic modelling, specifically, how to represent biological concepts like "byproduct burdens" and nutrient exchanges mathematically so a computer can optimise them.
What's next for biox-biohackathon-2026
Our immediate next step is to connect our offline optimisation logic directly to the live web interface, replacing our placeholder data with real-time, automated diet suggestions. We also have artificial intelligence (LLM) features built into the code but currently disabled; our goal is to turn these on so the platform can explain its biological findings in plain English. Finally, we plan to package the whole project into a Docker container so other scientists can install and run it with a single click.
Log in or sign up for Devpost to join the conversation.