The Great Vampire Squid of Healthcare

When we learn about incredible AI breakthroughs, it can be hard to figure out what it all means.

What are the practical implications? And how will this kind of breakthrough research lead to new products and investment opportunities?

Yesterday, we explored how DeepMind, a division of Google, essentially cracked the code on life over the course of four years. Its use of deep learning – a form of artificial intelligence – to accurately predict how proteins fold… worked.

And if that wasn’t a big enough accomplishment, AlphaFold 3 exceeded all expectations – by not only predicting the structure for all known proteins, but also that of DNA, RNA, and even ligands (a substance that binds with a biological molecule, like binding to a protein).

As I noted in yesterday’s Bleeding Edge – Can an AI Win the Nobel Prize, this is like having the key to all biology.

This breakthrough is what resulted in two DeepMind executives – Demis Hassabis and John Jumper – winning the Nobel Prize in Chemistry.

These developments are only months old. They’re so new and exciting, and there is so much information available. The life sciences community is digging in and looking for ways to leverage this information in their own research and development.

The future implications are hard for anyone to grok.

For perspective, more than 2 million researchers in more than 190 countries are already using AlphaFold data, developed by artificial intelligence, for their own research and drug development. This is yet another simple example of how widespread the use of artificial intelligence is today – and it’s still just the tip of the iceberg.

The excitement around AlphaFold this year was so great that many missed another incredible development out of DeepMind last month…

Hello, AlphaProteo

As if its breakthroughs in life sciences weren’t enough already, DeepMind just released AlphaProteo – an entirely new artificial intelligence that designs proteins that can bind to target molecules.

A simple way to think of AlphaProteo is as the applied science counterpart to AlphaFold.

I said that AlphaFold 3 is the key to all biology because all biological processes in plants, animals, and humans depend on interactions between proteins. They are the building blocks of life itself.

Proteins bind to other proteins in ways that regulate biological processes. If there is a bad protein in our bodies like the spike protein created by COVID-19, our immune system produces antibodies (a protein) to bind to the spike protein and protect the human body.

However, our bodies don’t always have a solution to fight against certain proteins associated with disease. This is the purpose of AlphaProteo – to design novel proteins with remarkable binding affinities for target proteins to cure disease.

The key was AlphaFold understanding how all proteins folded.

Training on AlphaFold

If we understand the structure of a protein, then we can use AI to model and determine how that protein interacts with other molecules in the body, as well as how a drug might be designed to regulate that protein that is causing trouble.

AlphaProteo Designed Binder | Source: Google DeepMind

The team at DeepMind took AlphaFold 3’s database, as well as protein data from the Protein Data Bank, and used that to train a new artificial intelligence (AI) – AlphaProteo – capable of designing entirely new molecules that bind to specific areas on target proteins. An example is shown above.

This first generation of AlphaProteo has already demonstrated astounding results, achieving anywhere from 3 to 300 times better binding affinities than any other method that exists today.

DeepMind chose seven target proteins to test, two of which are viral – BHRF1 and SARS-CoV-2 spike protein – and five others that are related to cancer, inflammation, and autoimmune diseases – IL-7Ra, PD-L1, TrkA, IL-17A, and VEGF-A.

AlphaProteo Target Proteins | Source: Google DeepMind

Once DeepMind developed binders for its target proteins, it collaborated with the Francis Crick Institute, which has a wet lab where the newly AI-designed molecules from AlphaProteo can be produced and tested in real life.

Here are the results:

Binding Affinity of AlphaProteo Binders | Source: Google DeepMind

Two things are important to note in the chart above.

The AlphaProteo data is in blue, and K_D (the Y-axis) is a dissociation constant, which means that higher numbers imply worse binding affinity.

Said another way, the lower the number, the better the binding.

The second point is that the Y-axis – which, again, is the K_D – is on a logarithmic scale, so the binding affinity of AlphaProteo binders is very impressive.

This is already a remarkable tool that can be used for drug development, the most obvious application of an AI like AlphaProteo.

But it can also be used to make crops more resilient, and even improve crop yields. It can also be used to create compounds that can be used to clean up an environmental mess.

These are all products that can be commercialized and invested in. And companies that use AI like AlphaFold and AlphaProteo are empowered to accelerate their product development.

Some of us might be thinking… to what end?

Why is Alphabet (Google) involved in life science research like this?

It’s an advertising company after all…

It’s true, of course, that most of Google’s revenue comes from advertising.

But it’s in the business of data collection and aggregation. The more data it can collect on every person on the planet, the better its ability to sell access to that data to advertisers.

Which brings us to the healthcare sector.

Google’s Closed-Door Play

Google Health has been developing what it calls its Articulate Medical Intelligence Explorer (AMIE), a generative AI designed to help clinicians understand medical information, diagnose conditions, and recommend a path for treatment.

Related to this initiative is a collaboration with Fitbit, a company that Google acquired for $2.1 billion in 2021, to design a health-centric large language model (LLM) to provide personalized health coaching based on each user’s own health-related data.

This kind of information was something that Google would have never had access to. But with this kind of personalized AI, Google can provide contextually relevant information to consumers with valuable health coaching that could be of great benefit to users.

Google will have the ability to deeply understand a user’s state of health. And of course, which products and services a person will benefit from – this is where advertising comes into play.

And there’s an even larger play by Google happening in the background, literally behind closed doors.

Very few people know this, but Google is sitting on roughly a $2 trillion portfolio of investments in both private and public companies. It is one of the most prolific corporate investors in history.

And, as I’m sure you’ve guessed by now, one of its most concentrated areas for investment has been in biotech and healthcare.

And most of those critical investments are held within a venture capital firm known just as GV, short for Google Ventures, which is entirely backed by Google and no one else.

More than half of the investment team at GV is focused entirely on biotech and healthcare.

For anyone curious about the incredible span of investments, and influence, that GV has in the industry, you can go here to see GV’s portfolio of investments.

Some notable investments in biotech are Revolution Medicines, Beam Therapeutics, Sana Biotechnology, Verve Therapeutics, Prime Medicine, Oscar Health, and many more…

There are too many to mention, its hooks are everywhere.

I suspect that Google has already made more profit off of its healthcare-related investments than the amount that it has invested in research and development in this space. But that’s not where the really big money is…

Google could certainly license access to its technology to the biopharmaceutical and life sciences industry, but I doubt it will.

Just like it gives free access to its search engine – to drive adoption.

It will likely pursue the same strategy in life sciences…

All in exchange for the right to collect more data, which can be monetized further down the road.

And that is another trillion-dollar opportunity.