02670nas a2200385 4500000000100000000000100001008004100002260001500043653001500058653001700073653004500090653001700135100001900152700001800171700001800189700002100207700002200228700002400250700002300274700002100297700001600318700001800334700001500352700002200367700001800389700001700407700001700424700001700441245011200458856005500570300000700625490000600632520163200638022001402270 2026 d c2026-02-0610aBiophysics10aCell biology10aComputational biology and bioinformatics10aNeuroscience1 aDavide Ortolan1 aPushkar Sathe1 aAndrei Volkov1 aDominik Reichert1 aSheldon Sebastian1 aArvydas Maminishkis1 aNicholas J. Schaub1 aBengt Ljungquist1 aDevika Bose1 aJorge Ferrari1 aNyusha Lin1 aGianluca Pegoraro1 aCarl G. Simon1 aRuchi Sharma1 aPeter Bajcsy1 aKapil Bharti00aAI driven 3D subcellular RPE map discovers cell state transitions in establishment of apical-basal polarity uhttps://www.nature.com/articles/s44387-026-00074-6 a200 v23 aThe retinal pigment epithelium (RPE) is a specialized cell monolayer that forms the barrier between the subretinal and choroidal spaces. During development, RPE cells polarize perpendicular to the monolayer plane such that organelles attain specific intracellular locations. This allows the RPE to differentially interact with overlying photoreceptors and underlying choriocapillaris. When RPE polarity is disrupted, tissue homeostasis is disturbed, leading to retinal degeneration. The subcellular organizational principles of RPE polarity are unknown. We developed an artificial intelligence (AI), specifically a mask region-based convolutional neural network-assisted high-content image analysis platform combined with mathematical modeling to develop a quantitative three-dimensional digital twin of RPE subcellular structures during the establishment of apical/basal polarity, polarity organization with learning-based analysis for RPE image segmentation (POLARIS). We discovered, during apical/basal polarization, cells constrict along the lateral axis and elongate apically, nuclear volume decreases, nuclear envelope develops invaginations, junctional complexes consolidate to the lateral membrane, the endoplasmic reticulum and mitochondria increase in volume and translocate towards the nucleus, and lysosomes move towards the central-apical side. AI algorithm and mathematical analysis reveal non-stochastic cell state transitions and organelle interactions in 3D during RPE polarization. These integrated AI-based quantitative data provide a reference digital twin to discover intracellular defects in diseased RPE. a3005-1460