AI

Alphabet’s Isomorphic Labs reportedly raising $2B+ for its medical research AI

Isomorphic Labs Inc., an Alphabet Inc. unit using artificial intelligence to speed up drug development, is reportedly in talks to raise a new funding round.

Sources told Bloomberg today that the investment could be worth more than $2 billion. OpenAI Group PBC backer Thrive Capital, an existing Isomorphic Labs investor, is expected to lead the round. Alphabet will reportedly chip in as well. 

Isomorphic Labs spun off from the search giant’s Google DeepMind machine learning research lab in 2021. It’s led by DeepMind co-founder Demis Hassabis. The executive and another DeepMind staffer jointly won half the 2024 Nobel Prize in chemistry for their work on an AI model called AlphaFold 2. The algorithm can automatically predict the structure of proteins, a task that historically required years of manual work.

In 2024, DeepMind and Isomorphic Labs released a significantly improved successor to the model. AlphaFold 3 delivers a more than 50% accuracy increase across certain research tasks. It’s particularly adept at analyzing small molecules, biological building blocks that form the basis of many drugs.

Isomorphic Labs’ flagship product is an AI system called the Isomorphic Labs Drug Design Engine, or IsoDDE. It’s designed to speed up drug development projects by automating manual research workflows. The Alphabet unit says that IsoDDE can perform some tasks significantly better than Alpha Fold 3.

Researchers often start a drug development project by studying the proteins that make up a disease-causing cell. In particular, they search for proteins with so-called binding pockets. Those are entry points through which drugs can be delivered into the disease-causing cell to neutralize it.

Isomorphic Labs evaluated IsoDDE’s ability to find binding pockets and perform related tasks using a benchmark called Runs N’ Poses. According to the company, the model more than doubled the score of AlphaFold 3 across the most difficult tasks. Another selling point of IsoDDE is that it requires a relatively limited amount of data about a protein to search for binding pockets. That reduces the amount of preparatory work involved in pharmaceutical projects.

After researchers identify a suitable binding pocket in a protein, they have to find a therapeutic compound that can attach to it. IsoDDE speeds up that task as well. It can predict a compound’s binding affinity, or its ability to attach to a protein in a disease-causing cell, with significantly higher accuracy than earlier methods.

Isomorphic Labs reportedly plans to use the funding round that it’s raising to enhance IsoDDE. The Alphabet unit also plans to grow its international presence. 

Image: Isomorphic Labs