Early Frames to Full Picture: Deep Learning for Interpretable mRNA-LNP Delivery Prediction

Overview

Lipid nanoparticles (LNPs) are an important component of a leading drug delivery system. They protect the medicine they carry, increasing its bioavailability and stability and notably used in COVID-19 vaccines and other mRNA vaccines that could treat solid tumors and personalize cancer vaccines. As testing the success of drug delivery via LNPs is notoriously time consuming, a novel machine learning CNN-Transformer hybrid architecture was developed that learns to predict the final result from just the first 3.5 hours of microscope images of target cells, instead of the full 12 hours of images, letting researchers test 3 to 4 times more medicine designs each day. The model works 14% better than the best published work. A Grad-CAM heatmap is generated revealing which part of the cell the model is examining when predicting. This could accelerate the development of cancer treatments, gene therapies, and vaccines for future disease outbreaks.

Awards (1)

• Selected for CWSF 2026

Competition history

• CWSF 2026 Disease & Illness

Related projects

Closest projects by meaning, across every fair and year in the corpus.