Inspiration
The COVID-19 pandemic showed us how quickly a virus can mutate and outsmart our defenses. ViruScope was created from the need to predict viral evolution before outbreaks strikes.
What it does
ViruScope predicts future influenza strains using AI. It analyzes past genomic data, spots evolutionary trends, and gives scientists critical lead time for vaccine development.
How we built it
We built a pipeline that cleans raw FASTA data, extracts key genetic features using bioinformatics algorithms, and visualizes insights in a clear, user-friendly interface. The code reads a large metadata CSV (which includes collection dates for each virus sequence), parses each date flexibly so we can sort all sequences from oldest to newest. Next, it selects the earliest sequence as a baseline for comparison. Each new sequence is pairwise-aligned with this baseline at the codon level, letting us label which codons changed. Those labels become targets for a Graph Neural Network (GNN) in which each codon is a node. Finally, the GNN uses two output heads—one predicting mutation probability (binary) and another predicting which codon each site becomes (64‐class). Through this approach, we clean the data, extract meaningful bioinformatics features, visualize the results, and maintain a minimalist homepage that boosts focus and user engagement.
Challenges we ran into
The dataset was too large to handle, so we had to reduce the FASTA file to only the most important sequences. This helped speed up processing while keeping key insights.
Accomplishments that we're proud of
We developed a robust AI model that transforms raw genomic data into actionable insights. This tool empowers proactive vaccine production and equitable health solutions.
What we learned
Integrating AI and genomics can revolutionize disease prevention. Simplicity in design and rigorous ethical safeguards are essential for meaningful innovation.
What's next for ViruScope
We aim to extend our predictions to other viruses and refine our model further. Our vision is a future where proactive global health defense is a reality.
Built With
- csv
- fatsa
- gnn
- neuralnetwork
- python
Log in or sign up for Devpost to join the conversation.