Fixing heart disease is like managing traffic in a city

In early 2021 our venture builder, Curiosity, co-founded Multiomic, a company that has a credible strategy to improve much of how we treat heart disease (or rather an even broader set of conditions called metabolic syndrome). And this will have a huge impact on our understanding of extending life.

I want to write a post about our thinking. This is aimed at a general audience so I will try to simplify everything as much as possible.

What exactly is metabolic syndrome?

Someone is diagnosed with metabolic syndrome if they exhibit 3 or more of excess blood glucose, hypertension, high triglycerides, high LDL cholesterol and excess weight. Most people with metabolic syndrome inevitably end up with much more serious conditions such as heart disease, type 2 diabetes, chronic kidney disease and fatty liver. These metabolic syndrome-related conditions (“MetSyn” for short) are — this is important – driven by the same underlying mixture of root causes.

It is the biggest cause of death worldwide. In the graph above, much of cardiovascular disease, diabetes, liver disease and kidney disease has metabolic syndrome at its core.

$1.9 trillion is spent on it a year today, and the forecast is $5.6 trillion by 2040.

The most important feature of metabolic syndrome

The most important thing to understand is that MetSyn is a disease that has no single simple linear cause. This is in contrast to cancer where a single genetic mutation can be a huge driver of disease; or to conditions like Huntington’s where a single mutation causes 100% of the disease.

MetSyn is instead a disease that I like comparing to traffic in a city — caused by a myriad of large and small factors, and requiring systemic change to address. In other words, it is a complex system.

Many contributors into MetSyn are well-known. Glucose/insulin dysregulation is a huge one. A sedentary lifestyle. Stress-related hormonal issues. Genetics. Behavior. Diet. These are all very intertwined and it is difficult to understand what really causes what. We just observe that the system behaves less efficiently, which leads to ever-greater issues that ultimately lead to death, with things like blindness, stroke paralysis, life on kidney dialysis along the way.

Our bodies constantly operate a large number of metabolic processes — where one chemical is turned into another and another in long chains of reactions. This is what powers life. As in any process, problems occur. This is another point when the traffic metaphor is apt — you cannot stop traffic since that would be killing the city. You have to manage the complex system better, which is likely to require a complex solution.

MetSyn is a critical driver of both COVID deaths and longevity

Interestingly enough, the vast majority of deaths from COVID-19 are in people with metabolic syndrome and related conditions. This is a major factor that is driving increased attention to MetSyn today.

Something personally interesting to me is that MetSyn also seems to be very connected to longevity. A hint of this is that the most validated life extension interventions — caloric restriction, metformin, rapamycin — are all highly connected to insulin regulation, which is at the absolute center of MetSyn.

So since I want to help drive significant progress in life extension over the next 30–50 years, working to fix MetSyn is a very interesting area of focus.

The need for data

The best technological tool we have developed to understand and affect non-linear complex systems is blackbox supervised machine learning. Modern neural networks can deduce relationships a human expert would never notice, represent health or disease states as combinations of a large number of factors, and rapidly classify whether a particular intervention makes things better or worse.

The challenge is this requires a lot of training data. And in this particular case, the data has to be:

1. Covering very many biochemical datapoints per person — what is called “multiomics.” You have to see deep details about the person’s DNA, RNA, proteins and metabolites (if you don’t know the relationships between these I highly recommend learning about it).
2. Phenotypically labeled — you have to know what is going on with the person, i.e. their sickness/health states as recorded by medical professionals in their health records
3. Longitudinal — many data points per person over time for both multiomics and phenotypes since the condition constantly evolves.

If we have such data, we will be able to deduce very detailed high-dimensional biomarkers of health/disease just by taking a blood (and/or urine/saliva) sample.

Wait, don’t we already have biomarkers for metabolic syndrome?

HbA1C, triglycerides, LDL-P, HDL-P etc. are all biomarkers which are definitively linked to MetSyn. The problem is they are too simple and clearly only explain some small part of the picture correlated to MetSyn.

So we give people drugs that reduce these biomarkers, but reducing these biomarkers has limited results, sometimes no results. It is as if we realize that a particular street in the city is highly correlated with overall traffic, so we expand that street hugely and now there’s no traffic there, but the rest of the city is still gridlocked.

The biomarkers we have now highlight some small part of a much larger and much more complex problem.

What is nice about the existence of low-dimensional biomarkers — and that new low-dimensional biomarkers are constantly discovered — is that they are a reason why we can be confident that high-dimensional biomarkers are there to be found.

High-dimensional biomarkers are key enablers of treatment

If we had a high-dimensional black-box neural network which described whether someone has lesser or greater metabolic disease, and which captured a large part of the explanatory power (unlike say HbA1C) then we could do very powerful things:

1. Develop treatments much faster by optimizing the high-dimensional biomarker rather than wait for people to not have heart attacks
2. Identify sub-clusters of people likely to respond to one treatment vs another treatment — what is called personalized medicine

Fascinatingly enough, personalized treatments for metabolic syndrome are close to non-existent despite the fact that it is the world’s largest disease. Example: because I have a genetic predisposition to heart disease as a result of an elevated Lp(a) biomarker, my doc Peter Attia and I decided I should be taking statins long-term. We tried Lipitor and it did not reduce my LDL-P. Then we tried Crestor and it did, so we stuck with that. Does this actually reduce my risks? It seems possible, and it has little downside, so we factor in the asymmetric risk-return profile and just do it. But we don’t really know. And there is no test that can tell in advance who will benefit from which treatment.

What we are doing at Multiomic

Our goal is to develop high-dimensional biomarkers for metabolic syndrome, and use these as a base to develop new drugs, segment the population to who is likely to respond to what intervention, and make ongoing treatment decisions. The vision is to build a metabolic disease focused full-stack pharma company that is based on a data-first culture and creates various flywheels to constantly gather more data.

That’s the long-term. The starting point is that we are partnering up with a number of hospitals, biobanks and startups to get biomaterials and medical records, run genomic, transcriptomic, proteomic and metabolomic tests on these biomaterials, co-develop classifier systems on the results, run commercial partnerships, and use the results to gather more data and make our classifier systems more powerful.

It is a really cool team — co-founder and CEO Robert is a 30-year veteran of life sciences business partnerships; co-founder Alasdair worked as a data exec at Bayer and bio VC at Index Ventures and Mubadala; COO Slaven is a Forbes 30-under-30 PhD with work in longevity; Head of Ops Celia ran clinical trials at Novartis; Head of Computational Bio Ariella led data science at a $2bn biotech in Boston; Head of Disease Bio Cristina spent over 20 years initiating and leading new drug discovery programs. Plus we have already raised two funding rounds from great investors who understand both tech/data/ML and biology.

We are constantly resolving very interesting technical, legal and business challenges, so it is a rapid learning curve. When dealing with biology we have to consider many nuances which I have not seen in mainstream tech — e.g. auditing whether samples in a hospital are comparable to each other despite being stored in different refrigerators is a particularly fun one.

We have a shot at figuring this out, and it is an opportunity that is ridiculously large in scale plus an approach that can work. If this is interesting, get in touch :)