Researchers at the National Institutes of Health (NIH) say they have developed a highly detailed digital replica of key cells in the human eye, which may advance the study of vision-threatening diseases such as age-related macular degeneration (AMD).

The research was funded by the NIH/NEI Intramural Research Program and was published in Nature Partner Journal–AI.¹

The new platform, described as a “digital twin,” allows scientists to study how retinal pigment epithelial (RPE) cells organize themselves in healthy and diseased states at unprecedented subcellular resolution. The researchers says the work provides a powerful new framework for understanding disease mechanisms and accelerating therapeutic discovery.

“This work represents the first ever subcellular resolution digital twin of a differentiated human primary cell,” said Kapil Bharti, PhD, scientific director at the NIH’s National Eye Institute (NEI). “It demonstrates how the eye is an ideal proving ground for developing methods that could be used more generally in biomedical research.”

RPE cells play a vital role in maintaining vision by supporting and recycling components of light-sensing photoreceptor cells in the retina. In diseases such as AMD, RPE cells degenerate, leading to the subsequent death of photoreceptors and irreversible vision loss. To function properly, RPE cells must maintain a precise top-to-bottom organization, known as apical-basal polarity. The apical surface faces the photoreceptors and is responsible for recycling worn-out cellular components, while the basal surface interfaces with the blood supply to deliver nutrients and oxygen and remove waste.

To model this complex architecture, NIH researchers created a three-dimensional, data-driven digital twin of RPE cells derived from induced pluripotent stem cells (iPSCs). The stem cells were developed by the Allen Institute for Cell Science in Seattle, and the RPE cells were generated at NEI. Using an automated confocal microscope, the team collected 3D imaging data from approximately 1.3 million RPE cells across nearly 4,000 fields of view.

The massive dataset was used to train an artificial intelligence algorithm called POLARIS—short for polarity organization with learning-based analysis for RPE image segmentation. The AI system was designed to identify cell structures such as the nucleus, measure cell shape and volume, and generate detailed 3D segmentations across different stages of cell development.

Researchers focused closely on how polarity emerges, quantifying the size, shape, and spatial organization of organelles and cytoskeletal structures as cells matured. Their analysis revealed that healthy RPE cells follow a predictable and orderly path toward a fully polarized state.

The result is an AI-driven atlas of polarized and non-polarized RPE cells that serves as a reference map for investigating how diseases disrupt cell organization at both the cellular and subcellular levels.

“The digital twin approach represents a powerful new tool for AMD therapeutic development and could be adapted to study other eye and non-eye diseases and conditions affecting cell polarity,” Dr. Bharti said.

“By combining AI with mathematical modeling, we’ve created a window into cellular processes that were previously hidden from view,” said Davide Ortolan, PhD, NEI research fellow and the study’s first and senior author. “This technology doesn’t just help us understand what’s happening in AMD—it gives us a platform to discover how to fix it.”

Reference

1. Ortolan D, Sathe P, Volkov A, et al. AI driven 3D subcellular RPE map discovers cell state transitions in establishment of apical-basal polarity. Nat Partner J AI. 2026. doi:10.1038/s44387-026-00074-6