Avatar Dragons

You can put a name to a disease with modern medicine, but when it comes to why one patient and another with the exact same diagnosis don’t get the same result from the same drug, we are not as good at giving you an answer.

Nowhere is this more apparent than in the world of immune-based medicine. Take cancer: you might have one person who has a remarkable reaction to checkpoint inhibitors, CAR-T, or a vaccine, and then you have someone with the same kind of cancer for whom those very things do little. Or in autoimmunity, where one patient will do well on a TNF inhibitor and another has to go through a round of different biologics before they hit on something that does the job.

It’s not for want of powerful therapies. The issue is the immune system itself: it is complicated, ever-changing, and unique to the individual. A label for a disease doesn’t always show you what is going on under the hood.

We set out to address this with ImmunoMap.

Think of it as an AI-driven way to profile a patient’s immune “failure modes.” We want to move past the question of “What is the patient’s disease?” to “What is the immune system doing wrong here, and how do we treat it?” In short, we want to make some of the guesswork in medicine a thing of the past.

The Core Hypothesis

Our view is that a lot of treatments fall short because they are picked by the book of the disease, not by the state of the immune system.

In oncology, there are any number of ways an immunotherapy can be thwarted. The tumor could be hiding from the immune system. Or maybe the immune cells see it but can’t get in. T cells may make it in and then just give up. The tumor can even shed the very antigen you’re after. Then there’s the risk of over-activating the system and causing toxicity.

These are distinct issues. You can’t fix them all with the same tool. If a tumor is “cold,” simply taking off the brakes won’t cut it. If it has strong neoantigens, a custom vaccine is a better call. With a clear surface target, you might be better off with a cell therapy or bispecific. And if the tumor is all over the map, you need a combination.

The same holds for conditions like psoriatic arthritis or IBD. The symptoms may check the box for a diagnosis, but the pathway driving it could be anything from TNF to IL-17, JAK-STAT, or even the microbiome. ImmunoMap is built on the notion that you should start with the mechanism, not the label.

How It Works

ImmunoMap would put together a full picture of a patient’s immune state from several sources.

First, the clinical side: the history, the scans, the flares and remissions, what has been tried and how it was received. The immune system leaves its mark in ways a physician can observe.

Then we look at the molecular level. For a cancer patient, that means looking at mutations, gene expression, and so on. For an autoimmune case, we’d be looking at cytokines, metabolomic markers, and the like.

But blood work only tells part of the story. You need to see what is happening in the tissue — be it a tumor, a joint, or the gut. Are the immune cells in there or being kept out? Is there fibrosis or local infiltration? That is the third piece of the puzzle.

Finally, you have to account for time. A tumor will change after you treat it; an autoimmune condition will ebb and flow. What you see as a biomarker today might not be relevant in six months. So the platform would be updated with new information as it comes in.

What you get from ImmunoMap isn’t a simple yes or no. It’s a map. In the case of a cancer, for instance, we might put a tumor in a category like “exhaustion-dominant” or “high-toxicity risk” and point you to the treatment that makes the most sense biologically.

With an autoimmune patient, the platform can put them in a box: which inflammatory pathway is in charge, what tissue is involved, how much of a flare we’re looking at, biomarker activity, and where they fit in terms of treatment response.

We’re not out to put doctors out of a job. The point is to arm the clinician with a more transparent read on the immune system before they make a call on therapy.

Immune Failure Modes

At its core, ImmunoMap is based on the premise that you can put a name to how the immune system fails.

Take the “invisible target.” In cancer, the T cells don’t see the tumor for what it is because there are no strong antigens to go on. With autoimmunity, the mechanism at play might be hard to spot from a routine blood work or the way the patient is presenting.

Then there’s immune exclusion. You have the right cells, but they can’t get in. Solid tumors are the poster child for this; between the density of the tissue, the odd vascularity, low oxygen and all the suppressive signals, it’s a wall for the immune system.

Or you have suppression. The cells make it to the site, but something in the environment puts the brakes on. In a tumor, you could be up against regulatory T cells, checkpoint pathways, macrophages, or even metabolic stress.

Exhaustion is another one. The T cell has the target in its sights but just can’t keep up the work. That’s often why a drug has an effect at first and then tapers off.

Overactive inflammation is self-explanatory, and it’s the flip side of the coin in autoimmunity: the issue isn’t a lack of immunity, but too much of it, and in the wrong direction.

Heterogeneity is also a factor. A disease doesn’t always run on one track; different areas can be driven by different things, which makes a one-size-fits-all approach less effective.

And finally, escape. Put enough pressure on a disease and it will find a way around it. Cancer cells can shed their target antigen, or an inflammatory route can be bypassed if you block one. You have to plan for resistance if you want to treat long-term.

All of this is to get us past the era of one-marker medicine. A single biomarker has its place, but the immune system is a system. We couldn’t have done this kind of thing twenty years ago; we needed the technology to catch up.

Now we have it. Tumor and RNA sequencing to find mutations and active pathways. Proteomics and metabolomics to see the proteins and small molecules at work. Spatial transcriptomics to put the immune cells on a map within the tissue. Radiomics and machine learning to spot patterns in the data that you can’t see with the naked eye.

In oncology, they’ve already started to move on. PD-L1, mutational burden, microsatellite instability — they’re helpful, but you need more than that. The new wave is to bring in genomics, the microbiome, imaging, and clinical data to make better calls.

Autoimmunity is following suit. There is a push to use biomarkers to get to a diagnosis sooner, to be more exact with who gets what. This is where we need our three expert validators to come in.

For ImmunoMap, we’re after feasibility: is the science sound, do the biomarkers hold water, and will a doctor put his or her trust in it?

We’ve brought in Francesco Ciccia, Vinod Chandran, and Robert Inman.

Ciccia is the man for the tissue-immunology end of it. Given his background in rheumatology and systemic disease, he can tell us if these failure modes make sense in the real world. We want to know from him if we’re oversimplifying things, or if a tissue-level profile is a must.

Chandran is our validator for the multi-omics and biomarker side. He’s been in the trenches with psoriatic disease, working on everything from proteomics to early prognosis. Since ImmunoMap is predicated on the notion that you can use biomarkers to sort patients and pick the right treatment, he’s the one to ask: are these markers up to the task? Can you see the tissue problem in the blood? What do we need to build a prototype that works?

You can’t make a first pass at ImmunoMap and expect to be all things to all cancers and autoimmune conditions. It’s too much of an ask.

So the MVP has to be more of a laser focus: one disease where we already know immune profiling and biomarkers matter. I see two good candidates to put in front of us: melanoma or psoriatic arthritis.

Take melanoma. It’s a no-brainer for a cancer use case since immunotherapy is the name of the game. There’s a lot of work being done on checkpoint inhibitors, neoantigens, and why some tumors resist treatment. We could build a version of ImmunoMap that puts a tumor in its place — whether it’s hot, cold, suppressed, or if the antigens are visible.

Or you have psoriatic arthritis. It’s a mess of a disease, hard to call early on, and tied to psoriasis. An MVP here would let you look at disease activity, who’s at risk for a flare, and how to pick a treatment path.

In either case, we’re talking about a dashboard that runs on what’s out there in the public domain. You put in your data — clinical, genomic, proteomic, imaging, whatever you have — and it gives you back an immune-state read with a confidence score and some logic to go with it.

Say you run a cancer case through it:

Immune infiltration: low
Checkpoint activity: moderate
Neoantigen visibility: high
Tumor microenvironment suppression: high
Suggested direction: Don’t just do a checkpoint blockade; prime it with a vaccine or oncolytic first.

An autoimmune run might look like this:

Inflammation: high
Tissue remodeling: moderate
TNF-pathway: not much to see
IL-17/IL-23: very relevant
Suggested direction: Have a conversation about a biologic for that pathway, don’t just ratchet up the dose.

Now, this isn’t to say the software is making the call. It’s to show the thinking behind it. The whole point is to take some of the guesswork out of it.

With a cancer patient, you can’t afford to let them go through a few failed therapies. Immunotherapy is costly, hard on the body and the mind. If we can tell you why a certain approach won’t fly, you’ve saved time and can put together a better plan.

For someone with an autoimmune condition, they might be on a merry-go-round of treatments for years. A way to see which pathways are in charge or where the disease is heading could mean less damage down the road and a better life.

Then there’s the research side. We could be smarter about how we put together clinical trials. Why not enroll by immune failure mode instead of just a diagnosis? If you have a drug meant to fix immune exclusion, test it on the patients who are actually excluded. That will give you a straighter line in your results.

That’s the rub: ImmunoMap is as much for designing a trial as it is for the patient.

But we have some hurdles. First is the data. Multi-omics is pricey and you never know what you’re going to get from one hospital to the next. Some folks will have a tissue sample, some will only have a blood draw.

Second, biology is complicated. The immune system is not static. A model that works on one set of patients may not hold up in a different population or at a later stage. We’ll have to validate that.

Third, you can’t be a black box. A doctor is not going to trust something that spits out “do this” without any reason. We have to be able to explain it: here are the biomarkers, here is the evidence, here is how sure we are.

Fourth is the question of equity. We don’t want a tool that only the well-heeled hospitals with top-shelf sequencing can use. We need a tiered approach so even with basic clinical and blood work, you can still get value.

And of course, you can’t have any of this without putting it through the wringer with regulators.

We see a future in which no patient is put on an immune-targeted therapy without first having an immune map.

Take cancer, for instance. An immune map can tell you if a tumor is in need of being activated, infiltrated, or if it’s time to let go of the brakes with a checkpoint inhibitor, target an antigen, or go with a combination of approaches. With an autoimmune condition, the same applies: you can see if the problem is a particular cytokine pathway, some tissue remodeling, the biology of a flare, or even inflammation tied to the microbiome.

And as we go, ImmunoMap will be more than a static document. We want it to be living. As fresh data from blood work, scans, biopsies and how a patient is responding to treatment comes in, the map updates. You have to be adaptive because the disease is. Cancer is always one step ahead; autoimmune disease has its ups and downs. The immune system is in a constant state of flux, influenced by everything from stress and age to what drugs you’re on. A set-in-stone plan won’t do.

So much of what we do is about navigation, not just making a prediction.

It’s a straightforward notion: before you decide on a course of action for an immune-related illness, you ought to know what kind of immune failure you’re up against.

We got to this point by looking at where cancer immunotherapy is heading — away from the sledgehammer of broad activation and into the fine art of engineering. But it’s also relevant to the world of autoimmunity, where there’s still too much of a “let’s try this and see” approach.

Then there’s the matter of validation. We have a plan to put the concept to the test with some heavy hitters. Francesco Ciccia will look at the tissue-immunology side of things. Vinod Chandran will see if the biomarkers and omics are feasible. Robert Inman will give us the reality check on the clinical and translational end. They’ll let us know if ImmunoMap is any good, where we need to focus, and what the best way to start is.

Don’t get me wrong, we’re not claiming AI is a silver bullet for immunology. Our working theory is more down to earth: if you can measure the patterns that drive these diseases and put them in a clear picture, you can make better calls on treatment.

The tools of tomorrow won’t be the ones that just ratchet up the immune system. They’ll be the ones that are wiser, safer and made for the individual. ImmunoMap is our way of getting there — a means to chart the terrain of immune failure before you make your move.

If you want to get started, then go ahead and watch this video, explaining this idea in detail.

Sources

• Francesco Ciccia professional profile / CV, University Hospital Luigi Vanvitelli and rheumatology clinical/research background.
• Vinod Chandran, University of Toronto Laboratory Medicine and Pathobiology profile.
• Vinod Chandran, UHN Research profile.
• Robert Inman, UHN Research profile.
• Robert Inman, University of Toronto Department of Immunology profile.
• Wang et al. “Artificial intelligence–enabled multi-omics biomarkers for predicting immune checkpoint inhibitor response and toxicity.” 2026.
• Mokhtari et al. “A Comprehensive Oncological Biomarker Framework Guiding Immunotherapy.” 2025.
• Le et al. “Single-cell multi-omics in cancer immunotherapy.” 2025.
• Liang et al. “Transcriptomic Models for Immunotherapy Response Prediction Show Limited Cross-cohort Generalisability.” 2026.
• Chandran et al. “Biomarkers in Psoriasis and Psoriatic Arthritis — Where Are We Now?” 2024.