Virtual cells built from 4D AI models and 'digital twins' could speed up drug discovery
Mitochondria—tiny structures that convert nutrients into energy—are often depicted as discrete kidney bean-shaped objects. But in reality, they form a dynamic, interconnected network throughout the entire cell, rapidly splitting and fusing as they're transported to where energy is needed most.
Researchers at the University of California San Diego have developed virtual cells using 4D AI models and digital twins to accelerate drug discovery. Mitochondria, often depicted as discrete kidney bean-shaped structures, are actually dynamic and interconnected networks throughout the cell that rapidly split and fuse as needed. These networks can serve as markers of disease or test new treatments.
To quantify the morphological changes in these networks, the researchers employed two approaches. The first approach used 4D lattice light-sheet microscopy to capture the movement of mitochondria and other structures in three dimensions over time. This technique was employed to train a deep-learning AI model called MitoSpace on 40,000 4D movies of drug-treated cells.
The model was able to predict cellular health based solely on the shape and movement of mitochondria, grouping drugs by mechanism with 75% accuracy compared to 56% accuracy when trained on flat 2D images commonly used in large-scale drug screens.
The second approach involved creating a physics-based digital twin of a living cell. This twin was built from a 4D movie by defining a set of rules about how its organelles behave and implementing those rules in a model. The researchers found that the digital twin's response to drugs closely matched that of real cells. They treated cancer cells with 25 different compounds known to perturb mitochondria and produced 40,000 single-cell 4D movies.
Using this library, MitoSpace learned to distinguish between the mitochondria of different cells and group them based on similar responses to drugs. The model could predict the energetic state of the cell based on the shape and movement of mitochondria across 26 drug conditions.
These studies, both published in Cell, could reduce dependence on time-consuming lab experiments and accelerate drug discovery for various diseases, including cancer, diabetes, Alzheimer's, and pediatric mitochondrial disorders. The researchers believe that virtual cells built from 4D AI models and digital twins could become a general-purpose tool in cell biology, saving experimental effort and accelerating research in disease treatment.
Written by urgent.news from Phys.org's reporting — not their text. Machine-written — may contain errors; check the original before relying on it.
