In parallel with recent developments in machine learning such as GPT-4, a group of scientists recently proposed using neural tissue itself as a computational substrate, carefully grown to replicate animal brain structures. After all, if AI is inspired by neurological systems, what better way to compute than a real neurological system? Collecting developments from the fields of computer science, electrical engineering, neuroscience, electrophysiology and pharmacology, the authors propose a new research initiative, which they call “organoid intelligence”.

OI is a collective effort to advance the use of brain organoids—tiny spherical masses of brain tissue grown from stem cells—for computing, drug research, and as a model for studying, on a small scale, how a complete brain might function. In other words, organoids offer a better understanding of the brain, and OI aims to use this knowledge to develop neuroscience computing systems that train on less data and with less power than silicon hardware.

The creation of organoids was made possible by two breakthroughs in bioengineering: induced pluripotent stem cells and 3D cell culture techniques.

Building on the existing field of neuromorphic computing, where the structure of neurons and the connections between them are studied and simulated in silicon architectures, OI extends the engineering analogy with the ability to directly program desired behavior into the activation activity of animal brain cell cultures.

The organelles are typically 500 microns in diameter, which is about the thickness of a nail. The researchers say that as organoids develop, the neurons that make up the organoids begin to connect in networks and patterns of activity that mimic the structures of different areas of the brain. The development of the field of organoids was made possible by two breakthroughs in bioengineering: induced pluripotent stem cells (iPSCs) and 3D cell culture techniques. iPSCs are stem cells capable of developing into any cell found in the body of an animal, which are created by converting an adult cell back into a stem cell. These induced stem cells are then biochemically incorporated into the specific neurons and glia needed to build that organoid. Newly developed 3D scaffolding techniques allow biologists to grow iPSC-derived neural tissues both vertically and horizontally, allowing the organelles to develop the interneuronal networks seen in animal brains. Scientists have been studying 2D cultures for decades, but single-layer tissues are not capable of turning into brain-like networks like organelles.

Networks make organoids a powerful model for understanding and potentially exploiting the dynamics of brain activity. Jens Schwamborn, professor of cell and developmental biology at the University of Luxembourg, uses organoids to study the development of neurological disorders such as Parkinson’s disease. “We repeated the key features of the pathology. We see the loss of dopaminergic neurons, we see the appearance of protein aggregates relevant to the disease,” said Schwamborn, whose laboratory developed the organoid model of Parkinson’s disease. These platforms allow them to study the development of Parkinson’s disease on a small scale in the context of a cellular network, which single-layer cultures cannot do: “This is a major advantage. We can see signs of illness that we know occur in patients but have so far been unable to replicate in the lab. Now we can finally do it.”

“We don’t teach cells how to do it. [Organoids] end with the organization of structures in the brain. I think that’s the strength: the computing power comes from this organization.”
— Alisson Muotri, University of California, San Diego

Just as organoids themselves are the product of advances in bioengineering, their utility as models of neurological function is the product of several other biochemical innovations—electrophysiology and microfluidics. Researchers can now control organoid development more reliably and precisely than they could even half a decade ago, and can use this specificity to design organoids that mimic the network structure and cellular composition of certain cortical and subcortical structures. Alisson Muotri, a professor of pediatrics and molecular medicine at the University of California, San Diego, believes these structures could give them brain tissue information processing capabilities. “In 3D you see all this extra organization that you don’t see in 2D. It’s genetic. We don’t teach cells how to do it. They end with the organization of structures in the brain. I think that’s the strength: the computing power comes from this organization.”

The presence of consistent, stable organoids also allows scientists to make meaningful measurements of neuronal activity within them. Multielectrode arrays (MEAs) are panels of tiny electrodes capable of measuring and stimulating the electrical activity of neurons near the surface of an organoid. Flexible MEAs that can wrap around the organoid mass are able to record data from the entire surface, and not just from the bottom layer of neurons in contact with the Petri dish. By analyzing these recordings, scientists can infer how all these neurons are talking to each other. Using a suite of signal processing techniques called causal modeling, researchers can map the connections between neurons that make up networks of organoid functional structure. These network maps can then be used to track how information is being processed by the developing mass of neural tissue.

The scientists speculate that by making populations of neurons within organoids respond consistently and predictably to given electrical inputs, they can turn organoid systems into organic processing units that can harness the apparent information processing capabilities of neural tissue to create flexible and powerful computing systems.

Cortical Labs, a Melbourne-based biotech startup, is launching Dishbrain, the first trainable neuroscience computing platform. The company aims to provide end users as a cloud service with programmable, single-layer 2D neural cultures that have already been shown to reliably learn digital input/output patterns, such as playing the classic pong video game. Brett Kagan, the company’s chief scientist, says the company plans to launch the service by the end of the year: “By the end of this year, we should have a beta system that people can use either through the cloud or through a partnership. with us for internal use, log in and be able to run the most basic environments,” he said.

Although similar organoid-on-chip computing systems are not yet available, the OI team is optimistic about the pace of their progress. Professor Muotri believes that within a decade we may see the development of organoid computing systems: “We may see a prototype in the next two to three years,” he said. “It will take 5 or 10 years for things to become more reproducible with all the right tools.”

The group’s study was recently published in the journal Frontiers in Neurology.

From articles on your site

Related articles online