How to Bring Organoid Cultivation into Industrial Era

Organoids are mini lab-grown organs that help test drugs without human trials. But their cultivation has long been a gamble — poor consistency and lack of standards have blocked their clinical use.

Now, a five-year effort by Professor Gao Huijun's team at Harbin Institute of Technology, working with Chinese biotech firm BGI Genomics, has produced an intelligent manufacturing system that grows highly consistent "blood island organoids." This turns guesstimated manual work into a standardized, automated production line, bringing organoid manufacturing into the industrial age.

Conquering cell clustering

The team's first major technical battle was eliminating randomness during cell seeding. In manual operations, cells clumped together or left empty spots, causing organoids to grow with huge shape and function differences. The team initially assumed the 100-micron needle's inner wall was the culprit, so they spent a full year testing various hydrophilic and hydrophobic coatings. Nothing worked — cells still aggregated chaotically.

Realizing the issue was control, not materials, they developed a single-cell compliant visual servo control method. Their first iteration focused solely on density control: A camera recognized cell counts and adjusted fluid pressure accordingly. This pushed the self-assembly success rate to 72 percent — but there it stubbornly stalled.

Through deeper observation, they discovered cells naturally migrate toward the center. So they added boundary constraint algorithms to let the machine detect edges and confine cells within defined zones. Yet results remained unstable.

The breakthrough came in 2023 when they integrated density, boundary and spatial distribution into a single closed-loop feedback system. Instead of squeezing cells, the AI gently guides each cell to its optimal position using real-time visual feedback. This triple-parameter control finally achieved 98 percent uniformity.

Cell consumption dropped from millions to just a few hundred per batch, and the system now produces 1,000 identical organoids simultaneously — a fourfold increase in efficiency and 7.3 times better biological consistency than human hands.

Seeing inside without destroying

With uniform organoids in hand, the next hurdle was quality assessment.Traditional methods required destroying the organoid to measure drug response — a one-time snapshot that couldn't track changes over days.

The team proposed a radical solution: injecting magnetic micro-nanorobots (each the size of a single cell) into the organoid, then manipulating them externally to probe internal mechanical properties.

But controlling a robot in a microscopic space demands an exceptionally stable magnetic field — a global engineering challenge. "Move the magnet by a hair, and the robot instantly veers off course," explained doctoral student Lin Chengqian. For two years, they iterated coil designs, recalculated magnetic circuits, and refined control algorithms.

In 2024, the team succeeded with an octupole electromagnetic coil system that generates a perfectly uniform field within the tiny chamber. This allows the robot to gently touch cell walls and translate displacement signals into mechanical data — measuring elasticity, stiffness, and structural integrity — all without harming the organoid. This turned a destructive "snapshot" into a multi-day "continuous health record" for the first time.

Simultaneously, they tackled another pain point: Sending samples to external labs for gene expression analysis took two weeks, by which time the biological state had already changed. Their solution was a dual-channel compressed sensing technology that captures both mechanical and genetic data at an ultra-low sampling rate (just six percent of conventional data).

Now, researchers get both datasets within two hours, on-site, boosting efficiency by over 10 times. This precise quantification replaces guesswork, providing reliable metrics for tumor grading and drug efficacy evaluation.

AI-driven automatic system

Despite breakthroughs in cultivation and sensing, the lab remained a chaotic battlefield. Temperature control, imaging, culturing, and detection were scattered across separate devices; researchers had to set alarms for midnight liquid changes; lens fogging from humidity fluctuations ruined experiments; and commercial AI couldn't remember protocols or connect to hardware.

From 2024 to 2025, the team partnered with Yongjiang Laboratory to build an integrated, unattended system. Software and hardware sub-teams worked in parallel. The software group digitized over a decade of experimental logs, extracting standard protocols to build a domain-specific AI agent from scratch. The hardware team reconfigured every device, rewriting communication protocols to bridge the gap between instruments.

After countless on-site debugging sessions, they succeeded. The proprietary AI agent now autonomously controls timing, coordinates robotic arms for liquid handling, adjusts magnetic fields and temperature, and manages imaging. The system operates on a closed-loop logic of perception, decision, control and feedback, solving chronic industry problems like timeline chaos and data memory loss.

The final integrated smart manufacturing instrument, unveiled in late 2025, represents a complete leap from manual craft to industrial-scale production. Its achievements have been published in Science Advances.

"This system doesn't alter how life develops; it provides the engineering stability needed to bridge basic research and clinical translation," said Lai Yiwei, lead scientist at BGI Genomics.

Their next goal? To push this technology into personalized medicine, making organoid-based drug screening a practical tool for patients, said Gao.

Source: Science and Technology Daily