Over the last decade, it has become increasingly clear that inflammation is a key driver of deadly diseases, including cancer and heart disease. During that same decade, researchers developed powerful omics-level technologies — such as genomics, transcriptomics and proteomics — but they weren’t designed to study the dynamic, complex changes that occur within human tissue during inflammation.
The immune system operates through complex signaling and recognition interactions between immune cells and living tissues and cells. Because of that complexity, learning how to measure immune function and immune cell interactions in living tissue is open white space — the blank areas on the map where no one has ventured yet.
At Biohub, we’re taking an engineering-forward approach to studying the immune system: developing completely new ways to measure inflammation. Our goal is to watch immune cells in action within the context of human tissue — to measure how cells are communicating with one another, how they’re recognizing each other, and how their function becomes dysregulated during inflammation.
We have an incredible, multidisciplinary team that is creating fit-for-purpose tools to measure things that have never been measured before. With this capability in hand, we can develop powerful AI models to predict immune system function.
Our overarching goal is intervention: How can we intervene when inflammation is just starting to get going, so that people don’t develop diseases that are the result of a lifetime of inflammation? If we can measure and then take action, the hope is that common diseases will become rare.
Here are five approaches we are taking to measure the immune system in real time:
1. Build live tissue omics
At Biohub, we’re developing and using mass spectrometry-based proteomic and metabolomic technologies to study live tissue noninvasively. Mass spec instrumentation is well-used and understood, but there has previously been no way to apply it to live tissue and watch inflammation in progress. We’re building that capability for the first time.
When we announced Biohub’s RFA to encourage others to join this effort to develop technologies for spatiotemporal omics in live tissue, I heard from many in the community saying, “I don’t understand this RFA. This is impossible.” That’s the point — it’s impossible now, but let’s make it possible.
If we can watch inflammation develop in real-time, we can identify the molecules that signal early inflammation. Then we’ll know, for example, which molecular signals indicate the start of inflammation. With that knowledge, doctors could put a patch on the arm of people at risk for inflammatory disorders and watch those markers over time. When inflammation starts to simmer, they can intervene with either lifestyle modifications or drug treatments to turn it around.
Also read: 4 ways protein monitoring could change the way we stay healthy
2. Develop functional genomics and immune cell recognition platforms
In addition to a live omics platform, our team has built a powerful functional genomics platform that allows us to CRISPR genes and then read out how those genes affect the properties of immune cells. That gives us great data to get cause and effect in immunology. Right now, we are using it to uncover key regulators of IL-17A, a pro-inflammatory cytokine, and the initiation of psoriasis, a chronic, inflammatory skin disease.
We are also developing a platform we call Decoder. Before an immune cell attacks another cell in the body, it recognizes it — like a lock fitting into a key. Using high-throughput T-cell/antigen screening and AI, this tool decodes those interactions to predict immune cell decision-making and subsequent responses.
3. Create multidimensional maps of inflammation
Overall, we’re combining these different technology platforms to map inflammation across many dimensions. The live tissue omics give us the dimensions of metabolome and proteome, while our sensing platforms allow us to watch inflammation markers with exquisite time resolution in real time within tissue.
We aim to gather multidimensional data across many parameters: space, time, proteome, metabolome, perturbation, cell-cell recognition, and cell-cell communication. When we layer and integrate these datasets, we’ll be able to decode complex cell-cell signalling networks, and identify actionable drivers of inflammatory diseases.
4. Gather it all into a powerful virtual immune system
We will feed all this multidimensional data into the Virtual Immune System, an ambitious effort from Biohub to build AI-based, predictive models of human immunity, to map and predict how immune cells change.
In autoimmune disease, immune cells attack the body for reasons we don’t understand — sometimes they attack the skin as in psoriasis, or the gastrointestinal tract as in inflammatory bowel disorder, or the nervous system during multiple sclerosis. We can tell those dysregulated immune cells started as normal immune cells, but how did they end up in a pathogenic state?
If we can generate all the data to show how an immune cell transforms from healthy to pathogenic, we can then develop a predictive model to go backwards, from dysregulated back to healthy.
The Virtual Immune System will also help us predict why immune cells recognize things that they shouldn’t and model how immune cells interact within tissues. The Virtual Immune System is going to help us go way above and beyond what we can do experimentally, and it is possible thanks to predictive AI models.
5. Integrate AI throughout
At Biohub, we’re building frontier AI for biology, and we use AI to accelerate everything we do. For example, in our functional genomics platform, we recently screened a large set of perturbations looking for drivers of psoriasis. We made a library of gene knockouts for every gene in human skin cells, and we generated a huge dataset of potential hits.
Instead of manually parsing that data to search for links to psoriasis, we built an AI interface — a fit-for-purpose large language model that took all the data, plus scientific literature, and helped us to parse potential hits. We used the LLM to analyze this dataset and prompted it to point us towards novel biology that nobody’s ever seen before that’s druggable — specifically genes that appeared to drive psoriasis that we could modulate with a small molecule. Finally, we used the AI model to find druggable hits where there’s already an FDA-approved therapy that hits that target.
That whole experiment — from data gathering to wet lab validation — took six months end-to-end. Traditionally, this type of effort can take five years, and many of the resulting hits are often wrong. This is one recent example of how we are building and using AI to accelerate everything that we do.
At Biohub, we function like a startup: Our teams seek out the hardest problems, and everyone runs full throttle at the problem as quickly as we can. Measuring and tracking inflammation in a new way is one of those challenges. If we measure things that have never been measured before, we can approach preventing or curing inflammatory diseases in new ways.
Shana Kelley is president of bioengineering and head of Biohub, Chicago. There, she leads a scientific grand challenge focused on building tools to monitor immune cells within tissues in real time, and ultimately steer the immune system away from tipping points that lead to inflammatory disorders.