New Breed of Computers May Evolve From Live Cells

Researchers at the National Science Foundation (NSF) are studying the potential to harness the computer skills of tiny groups of biological cells known as organoids.

Brains, whether human or animal, are known to be biological supercomputers performing, in some cases, billions of calculations per second using very little energy. But all biological organs—plants, hearts, livers—perform various calculations. And small collections of lab-grown biological cells known as organoids may one day perform calculations that benefit humans in a number of ways, including advancing medicine.

But first, scientists have to understand how organoids work, what data they share and how that data may be captured and put to use.

Enter the NSF.

Through the Emerging Frontiers in Research and Innovation (EFRI): Biocomputing through EnGINeering Organoid Intelligence (BEGIN OI) program, the foundation has invested $14 million in the study of organoid intelligence to foster fundamental and ethically responsible research and development of organoid intelligence systems while also broadening participation in biocomputing research, according to an NSF press release.

“NSF’s investment will lead to biological computing with superior power and efficiency by harnessing the mechanisms behind complex biological behavior for smart systems,” Susan Margulies, NSF assistant director for the Directorate for Engineering, said in the release. “Advances in biocomputing will open new opportunities for artificial intelligence, biotechnology and more sustainable computing.”

Organoids are tiny, 3D versions of engineered tissue that replicate important functions of natural tissue in vitro, or in an artificial environment outside the living body. Organoid intelligence is an emerging multidisciplinary field focused on developing novel biological computing systems that emulate the flexibility, robustness and efficiency of cells and organs.

The NSF has chosen seven interdisciplinary research projects to receive $2 million each. The seven teams will use organoid intelligence systems to address current limitations in artificial intelligence technologies and potentially revolutionize the capabilities of biological computing, according to the press release.

Steve Peretti, program director, Cellular and Biochemical Engineering, Division of Chemical and Bioengineering, Environmental and Transport Systems (CBET) at NSF, explained in a SIGNAL Media interview that any organ—a liver or kidney, for example—has “a bunch of special functions” that grew from a tiny undifferentiated cell.

“An organoid is a clump of cells that has begun to differentiate in a particular direction. There are brain organoids, liver organoids, retinal organoids, stomach, gut lining organoids. Essentially, it’s a group of cells that is starting to display some of the unique characteristics of the organ,” Peretti said. “So, if it’s a liver organoid, it’s starting to behave like a liver, a fully differentiated, mature liver.”

Organoid research represents a new frontier for science with potentially profound implications. Some future computers may not look like today’s “in silica” computers, where “you put a bunch of electronic information in, and a bunch of things happen and information comes out,” Peretti suggested.

“Truth of the matter is, every organ in our body is doing computations right now. Your pancreas is looking at your blood sugar levels; your bone marrow is looking at your red blood cell count. There are things that are going on: chemical signals, temperature signals, acidity, salinity. Most of what they do is local calculations. They respond to the environmental cues, the inputs, and they make an action. That’s the output. They’ve got a calculation in between,” he said.

And those calculations do not involve traditional computing’s 1s and 0s. “It’s a lot more complex because these organs are taking tens of data points per second, per millisecond, per whatever, and making an equally large number of responses per second. How do they do it? That’s what we’re trying to understand. We know that they do it really well or else none of us would be here,” he offered.

Organoids have proven useful to researchers before. For example, they have been used for toxicology studies, allowing researchers to detect potentially hazardous products without conducting studies on living animals. So, for example, if scientists need to know how a particular chemical might affect the liver, they can use a liver organoid. The response may not be perfectly comparable to that of a fully developed liver, but it is a “first-level proximate response,” Peretti said.

Two-dimensional cell groups also have been studied. “We have been very good at studying planar sheets of cells. It’s easy to get access to them. You can shine lights on them and look at how they respond. You can put microelectrodes in and see how they respond.   

So, it’s easy to measure and observe,” he said.

But organoids are a different animal, to use a clichéd pun. “Organoids are incredibly complex, already three-dimensional. There’s information and crosstalk between all of those different cells. There’s a lot of crosstalk between individual cells, and this group is talking to that group and so on,” he noted. “That can help inform our understanding of so many different things, not just diseases.”

Among the questions researchers hope to answer is what the signals emitted from organoids mean and how they can map the inputs that the cell groups are exposed to with the organoid output. That’s where modern artificial intelligence (AI) capabilities could prove useful because the inputs and outputs are simply too numerous for teams of humans to track conventionally. “We couldn’t have done this five years ago because we didn’t have the computational tools to interpret what was coming out,” Peretti observed.

Asked what an organoid-enabled biological computer might ultimately look like, Peretti theorized, or possibly joked, that it may be somewhat comparable to an “old Mac Cube thing.” He added, however, that one challenge is figuring out how to keep the organoids alive.

“It’s just going to have a whole lot more wires and tubes and things running in and out of it because the only signals are not going to be electrical. They’re going to be optical. They’re going to be chemicals,” he said. “We’ve got to put nutrients and things in to keep it alive so it can do its computing. We’re going to have to keep it at a happy temperature and so on. So, it could look like a box with just dozens of tubes and pipes and whatever running in and out, depending on how we’re interacting with the biological system.”

The director indicated it is too early to define all the ways organoid computing might prove useful, but part of the promise is to get away from the restrictions of electrical inputs and outputs. “We get to sample a lot of larger environmental cues because cells don’t just respond to one chemical. They respond to [more]. They don’t just respond to temperature, they respond to gradients of temperature, gradients of light, directions of light. They’re talking to each other in ways that—and a degree of complexity that—we don’t know how to wire between two ‘microchips,’” he said.

Explaining the potential in terms an English major can understand, Peretti helpfully dropped a reference to a fictional character. “Once we understand how to talk to—Dr. Dolittle, how to talk to the animal—we’re going to be able to ask it complex questions that maybe we don’t even know how to fully define. We’ll be able to put them in environments that we want to know things about, and it will sample everything that’s relevant, even if we haven’t thought of it potentially.”

In response to a question about organoid computers and quantum computers, Peretti noted that he is not a quantum expert but that organoid computing might complement quantum technology. “I don’t think that it will make [quantum computing] obsolete because there are quantum effects that probably can’t be replicated and that have intrinsic value for data security and for other applications that are beyond what a biocomputational system would be capable of doing.”

Like their conventional computing counterparts, organoids will likely be most powerful in networks. “When we want to solve enormous problems using in silica computation, we have networked a whole bunch of microchips that are doing a whole bunch of calculations. We have a network of calculators doing computations in parallel and series. If you talk to the investigators, they are talking about not just a single organoid or a single collection of cells. They are talking about networks of cells in large three-dimensional meshes where each organoid might be a node in a larger biocomputer.”

The BEGIN OI program includes three threads: biocomputing theory and modeling; biology-integrated culture maintenance and hardware systems; and ethical, legal and social implications. The first acknowledges that understanding the output from individual organoids or groups of organoids will require new tools, including AI, which will help analyze the data and determine which information is useful or simply noise. “If we’re asking [an organoid] to add 2 plus 2, we have to figure out where does it say ‘4’ as opposed to 5? Can we set up a translator so that we can send it information that it understands, and it sends us information that we can understand?”

The second thread, he said, is developing the chassis “that allows the system to actually live and be able to do its computing.” And the third requires the teams, from the first day, to include ethical research questions for their projects “because if you wait five years before you ask the first ethical question, you’ve already gone down a road that may be hard to turn from,” Peretti asserted.

Organoid and biological computing research is so new that scientists don’t yet know how the technology will ultimately be used, but the research topic exemplifies what the NSF tries to do, which is to “find exciting technology that needs a nudge” to address national needs, even though the specific applications of the technology might be difficult to predict. “We don’t specify that you have to do this particular calculation or that particular calculation. Do some calculations with this thing; show us what’s possible,” Peretti said. “We are trying to understand how these things work by seeing if we can get them to do A, B, C. In the simplest terms, if we can get an organoid to be exposed to, let’s say, blue light and yellow light and give us a green response, like, that would be crazy.”