AI and Drug Discovery - Isomorphic labs
Thesis
Proteins are the fundamental building blocks of life. All living organisms - big and small, plants and animals - are composed of millions of proteins that serve different purposes and perform functions essential to the survival of these organisms. To understand the function of these proteins it’s important to know about their shape (geometry) and structure. This 3D representation of proteins can help us predict their purpose and model their interactions with other substances.
Earlier, scientists would use methods like nuclear magnetic resonance and X-ray crystallography to determine these 3D structures. This relies on extensive trial and error, years of work and millions of dollars. So for decades, scientists have tried to predict a protein’s structure from its amino acids alone. This came to be known as the protein folding problem.
In 2020, Google Deepmind released Alphafold 2, their deep learning model to predict protein structure just from a sequence of amino acids, which showed such a high level of accuracy that most people considered the protein folding problem solved.
Now, in 2024, Google Deepmind with Isomorphic labs released Alphafold 3 which models the interactions of all of life’s molecules. Isomorphic labs is the private subsidiary of Alphabet that is poised to use these developments in AI from Deepmind to discover drugs and treatments to diseases at a scale and level of personalization that wasn’t possible before. This business could be worth north of $100 billion according to internal estimates.
Founding Story
Isomorphic Labs was founded in 2021 by Sir Demis Hassabis, who is also the founder and CEO of Deepmind.
Demis’s journey in AI started as a child when he achieved master status in the game of chess at age 13. It was his winnings from chess that helped him afford his first computer, using which he taught himself how to program. He went on to receive his education at the University of Cambridge in computer science. Post his undergrad, he worked as a game AI developer at Lionhead and eventually founded his own gaming company, Elixir Studios. After leaving the company, he returned to education and pursued a PhD in cognitive neuroscience from University college london. Then in 2010, Hassabis founded Deepmind, the British AI startup behind breakthroughs like AlphaGo (Go playing AI) and Alphafold.
Given the research output in computational biology coming out of Deepmind, Hassabis envisioned a sister company that would work on building this foundational research into real world solutions that could help cure diseases and provide personalized medicare on the back of AI systems like Alphafold. Imagine using AI to reinvent a drug based on highly accurate 3D structure predictions of the pathogen. Isomorphic labs was incorporated in Feb 2021 and then announced in Nov 2021.
Products
While the company is still very young and has significant progress to make before its products can be made commercially available, Isomorphic Labs managed to secure two deals worth approximately $3 billion with Eli Lilly & Company and Novartis AG to collaborate on R&D in drug discovery.
The deal with Eli Lilly & Company involves an upfront payment of $45 mil, with potential total deal value reaching $1.7 bil. Novartis will pay $37.5 mil upfront, with potential total deal value reaching $1.2 bil.
Recently, Isomorphic Labs collaborated with Deepmind to co-author the groundbreaking Alphafold 3, a model capable of predicting the interactions of all known molecules. Alphafold 3 is freely available on the Alphafold server for non-commercial research.
PS
I’m only human, a biological language model (maybe little better than that I can give myself more credit), and I may make mistakes in my research. Open to feedback in the issues section of this blog repository.