Aug. 18, 2026 — Biology doesn’t occur at one scale. Molecular interactions unfold in milliseconds and nanometers, whereas disease-associated change resembling in Alzheimer’s spreads throughout millimeters of mind tissue. Understanding complicated organic methods requires scientists to observe each — ideally on the identical time and with the identical pattern. Traditionally, this has meant shuttling samples between specialised devices, usually damaging organic context and slowing outcomes. There’s additionally a typical crux throughout microscopes: the nearer you take a look at dwelling tissue, the extra the picture blurs, and the extra element you seize, the extra overwhelming the ensuing information turns into.
5 imaging modes present complementary views of the identical dividing human retinal pigment epithelial cell (hTERT-RPE1). From high left, clockwise: Lattice Mild-Sheet Structured Illumination Microscopy (LLS-SIM), widefield, 3D-SIM, indirect illumination, lattice mild sheet. This reveals how completely different microscopy methods evaluate when imaging the identical cell, giving them a extra full image of what’s occurring inside. (Credit score: Fu, Liu, Milkie, Ruan et al., Nature Strategies, 2026)
A brand new instrument goals to deal with these issues — and has revealed a 3rd, arguably tougher problem that Berkeley Lab is uniquely positioned to deal with.
Researchers at Lawrence Berkeley National Laboratory (Berkeley Lab) and collaborating establishments have developed the Multimodal Optical Scope with Adaptive Imaging Correction (MOSAIC) — a reconfigurable microscope that consolidates greater than ten imaging methods into one compact instrument. It processes its huge datasets utilizing computational instruments developed at Berkeley Lab, funded by a Laboratory Directed Analysis and Improvement (LDRD) award and supported by the Perlmutter supercomputer on the Nationwide Vitality Analysis Scientific Computing Heart (NERSC).
Featured on the duvet of Nature Methods, MOSAIC permits scientists to trace organic processes throughout scales and evaluate imaging strategies on the identical pattern. It generates information at a tempo that’s pushing the boundaries of what biology can uncover.
“MOSAIC can generate as much as 4 terabytes of knowledge per hour — far past what standard processing workflows or human inspection can deal with,” mentioned Srigokul “Gokul” Upadhyayula, a college scientist within the Molecular Biophysics and Built-in Bioimaging Division and co-corresponding creator. “The microscope is just as helpful as our means to course of these information and extract organic that means from them. Berkeley Lab’s experience in high-performance computing and large-scale information evaluation is important to closing that hole.”
How MOSAIC Works
MOSAIC grew from the adaptive-optical lattice light-sheet microscope reported by Nobel laureate Eric Betzig, Upadhyayula, and their colleagues in 2018. That earlier system delivered distinctive efficiency however was so giant it occupied a 10-foot by 4-foot optical desk. As demand from the broader analysis group grew, the staff designed the MOSAIC to retain and increase these capabilities whereas lowering the instrument’s footprint.
“MOSAIC has been constructed over a dozen occasions elsewhere with over 50 analysis licenses already shared. We additionally created complete documentation on learn how to construct this instrument — suppose an IKEA-style instruction set geared in the direction of a scientist who has no deep optical experience however is keen to study.”
MOSAIC’s important innovation is that it may be rapidly reconfigured in two to 5 seconds to modify between a dozen distinct imaging modes. The staff designed a wise modular system the place the identical lasers, mirrors, cameras, and computational {hardware} serve a number of imaging features via a customized optical switching system. Critically, each a type of modes is enhanced with adaptive optics: a know-how borrowed from astronomers who developed it to sharpen photographs of distant stars blurred by Earth’s ambiance. This corrects blurring brought on by aberrations within the dwelling tissue itself.
“ Pattern-induced aberrations distort and redirect mild, lowering each sign and determination,” mentioned Upadhyayula. “Adaptive optics measures these distortions and corrects them. It’s like turning on the windshield wipers whereas driving within the rain: the knowledge is current all alongside, simply obscured.”
MOSAIC additionally depends on fluorescent molecules that enable biologists to mark particular mobile constructions and molecular actions in dwelling cells, quick and delicate light-sheet imaging that captures mobile dynamics with minimal stress or injury, and high-speed information switch infrastructure able to transferring and processing huge imaging datasets.
What Turns into Seen When You Clear the Windshield?
MOSAIC’s means to picture with minimal invasiveness at giant scales over lengthy durations has already enabled a number of experiments: monitoring single molecules in dwelling cells, observing organelle dynamics in growing zebrafish embryos, mapping neuronal structure in expanded human mind tissue from an individual with Alzheimer’s, and imaging neural exercise in dwell mouse brains. In that final software, adaptive optics correction revealed roughly 2.5 occasions extra detectable neural calcium occasions than imaging with out it — suggesting standard microscopy has been quietly undercounting mind exercise.
MOSAIC can be the instrument that powered a related study on Volumetric Imaging through Photochemical Sectioning (VIPS), printed in Science in 2025. That undertaking used MOSAIC and the computational instruments developed at Berkeley Lab to picture two full grownup mouse olfactory bulbs at nanoscale decision, producing roughly a petabyte of knowledge in roughly two weeks. Analyzing it took two years: an illustration of the hole between what these devices can see and what researchers can at the moment course of.
“ The bottleneck is not our means to amass the info,” mentioned Upadhyayula. “These microscopes can generate huge datasets at staggering charges. The important thing bottleneck is popping dense five-dimensional observations into organic understanding.”
Berkeley Lab’s contribution helps to focus on this hole.
Spherical-the-Clock Knowledge Assortment for Organic AI
Supported by an LDRD award, Eric Betzig and Upadhyayula’s group developed PetaKit5D, an open-source software program toolkit that may deal with MOSAIC’s terabyte-per-hour output in actual time and cuts processing prices by greater than an order of magnitude in comparison with earlier approaches. The staff additionally secured computing allocations on the Perlmutter supercomputer at NERSC to course of and visualize parts of the biggest datasets. However processing information effectively is just half the equation. The opposite half is producing sufficient of it — persistently, at scale, and of enough high quality — to coach the form of AI mannequin that would someday make sense of all of it.
At UC Berkeley, two MOSAIC devices now run across the clock, capturing the five-dimensional information — three spatial dimensions, time, and molecular identification — that can be wanted to coach a brand new state-of-the-art AI mannequin.
The information flowing from these devices already represents a basic shift in how biology might be practiced. For the primary time, researchers can watch in vivo biochemistry unfold inside cells dwelling inside their native tissues, inside a dwelling organism.
What comes subsequent, he believes, might be transformative: a imaginative and prescient language mannequin that causes natively over biology, connecting what it sees with molecular identification, experimental context, and prior organic data to find out which observations matter and which experiments ought to come subsequent.
“Linked to automated microscopes, pattern dealing with, and perturbation methods, that functionality might present the muse for self-driving organic laboratories — and essentially change the speed at which we are able to make discoveries,” mentioned Upadhyayula.
Source: Jenny Weger, Berkeley Lab
