Virtual Cells Built from 4D AI Models Could Revolutionize Drug Discovery for Cancer and Alzheimer's

Virtual Cells Transform Drug Discovery Through AI-Driven Approaches

Researchers at University of California San Diego have developed virtual cells—digital models that replicate dynamic biological processes—using two complementary AI-driven approaches published in Cell. These innovations leverage 4D lattice light-sheet microscopy to capture mitochondrial networks in three dimensions over time, moving beyond traditional flat snapshots. The breakthrough could accelerate drug discovery for cancer, diabetes, Alzheimer's, and pediatric mitochondrial disorders by reducing dependence on time-consuming laboratory experiments.1

1

2

Source: Newswise

MitoSpace: Deep-Learning Model Achieves 75% Accuracy

The first approach trained a deep-learning model called MitoSpace on 40,000 single-cell 4D movies of cancer cells treated with 25 different compounds. Unlike conventional AI models requiring manual labeling, MitoSpace discovered patterns independently, learning what distinguishes one cell's mitochondria from another. The model successfully grouped cells with similar responses and could predict cellular health solely from mitochondrial shape and movement across 26 drug conditions.1

1

When trained on 4D movies, MitoSpace distinguished between drugs and grouped them by mechanism with 75% accuracy, compared to only 56% accuracy using flat 2D images common in current large-scale drug screens. "For a century we have believed that mitochondrial form reflects function; this shows the relationship is strong enough that a model can learn it without ever being shown the answer," said corresponding author Johannes Schöneberg, PhD, Roger Tsien Chancellor's Faculty Fellow and associate professor in the Department of Pharmacology at UC San Diego School of Medicine.2

2

Digital Twin of a Living Cell Predicts Drug Responses

The second study created a physics-based digital twin of a real cancer cell by mapping mitochondrial positions and microtubule tracks using specialized image-analysis software. The team added motor proteins that transport mitochondria according to previously established rates, incorporating laws of motion until the virtual cell's behavior matched real cells. "We have built a physics‑based virtual cell and can compare it side‑by‑side to the actual 3D microscopy movie, something that has never been possible before," Schöneberg noted.1

1

To validate the model, researchers tested how mitochondria would respond when microtubules were partially broken down by nocodazole. Without changing parameters, the digital twin accurately reproduced the reduced motion and fusion-fission rates observed in drug-treated real cells. This capability to simulate mitochondrial behavior could enable testing of drug effects, disease mutations, or cellular engineering designs before conducting physical experiments.2

2

Related Stories

Stanford's AI models unlock cross-species cell biology insights from 112 million cells

Scientists Propose Global Initiative to Create AI-Powered Virtual Human Cell

Stanford's 37,000 AI Agents Identify Promising Lung Cancer Drug in Virtual Biotech Breakthrough

Stanford Researchers Create AI-Driven 'Virtual Scientists' to Accelerate Scientific Discovery

Computational Biology Advances Beyond 2D Imaging

"A cell is a four-dimensional object: it has depth and it never stops moving," Schöneberg explained. "Virtual cells need to be built on data that captures that fact." The 4D approach proves critical because mitochondria form dynamic, interconnected networks throughout cells, rapidly splitting and fusing as they transport to where energy is needed. These morphological changes serve as markers of disease and treatment efficacy, but traditional flat snapshots failed to capture this complexity.1

1

MitoSpace demonstrated remarkable adaptability by organizing drugs it hadn't encountered during training and sorting human lung organoid cells by developmental stage without retraining. This versatility suggests the deep-learning model could become a general-purpose tool in computational biology, potentially revealing new uses for existing medications while speeding discovery of novel treatments.2

2

Source: News-Medical

Implications for Multiple Disease Areas

Both virtual cell approaches target diseases where mitochondrial dysfunction plays a central role. The ability to predict cellular health from mitochondrial networks could transform how researchers screen compounds for cancer therapies, diabetes treatments, Alzheimer's interventions, and pediatric mitochondrial disorders. By reducing experimental effort and accelerating research timelines, these 4D AI models and digital twins represent a significant shift in drug discovery methodology, moving from reactive laboratory testing toward predictive computational biology that guides experimental design more efficiently.