Google, NVIDIA, Microsoft are all making efforts, AI drug development is a billion-dollar market!

Author: Li Xiaoyin

Source: Hard AI

More and more tech giants are increasing their bets on AI in healthcare.

Overnight, Google's DeepMind and its sister company Isomorphic Labs announced a major upgrade to their AI drug development model AlphaFold, stating that the latest version AlphaFold 3 can predict the structure of biological molecules such as proteins, DNA, RNA, and how they interact with each other.

Alphabet and Google CEO Sundar Pichai stated that currently, over 1.8 million researchers are using AlphaFold for protein prediction in vaccine development, cancer treatment, and other research projects.

In an interview, Isomorphic Labs CEO Demis Hassabis mentioned that artificial intelligence systems could potentially revolutionize medicine and create "enormous commercial value":

"I hope to achieve two things through Isomorphic: to build a business worth tens of billions of dollars, which I believe has the potential; and to bring incredible benefits to society and humanity."

Google is not the only one eyeing this track.

Currently, almost all AI tech giants have shown interest in the biopharmaceutical field, with Microsoft, Amazon, and even Salesforce conducting protein generation projects.

Recently, Kimberly Powell, Vice President of Healthcare at NVIDIA, stated in a media interview that healthcare will become NVIDIA's next "multi-billion dollar business," with NVIDIA aiming to provide chips, cloud infrastructure, and other tools to more biotech companies.

The next frontier application of AI technology?

NVIDIA's founder and CEO Jensen Huang has emphasized multiple times that digital biology will be the "next amazing disruptive technology."

As he mentioned, at NVIDIA's GTC 2024 conference in March this year, healthcare remained one of the main focuses, with life science-related events ranking first among all industries.

Over the past two years, NVIDIA's AI drug development platform BioNeMo's venture capital arm Nventures has invested most of its money in drug development projects. Data shows that out of Nventures' 19 investment transactions, 7 were invested in AI drug development startups.

Powell explained:

"Computers have already assisted in designing the first $2 trillion chip company. Why can't they help build the next trillion-dollar pharmaceutical company?"

Several other tech giants are also focusing on drug development. In just the past year, Salesforce launched the protein generation AI large model ProGen, Microsoft released a similar open-source model EvoDiff, Amazon released a protein folding tool for its AWS machine learning platform SageMaker, and reportedly even ByteDance is recruiting scientific and drug design teams This raises the question: What is the pharmaceutical value of AI technology?

Taking the protein field focused on by AlphaFold as an example.

Proteins are important components of human cell tissues and the main bearers of life activities. Each protein is composed of a series of amino acids, and the interactions between these amino acids and the external environment determine the protein's "folding" manner - which determines its final shape, and the shape of the protein is the basis for its functional execution.

Therefore, the most valuable point for technology companies is: if it is possible to predict the shape of proteins based on their amino acid sequences, they can be applied to drug development, crop improvement, biodegradable plastics, and other major fields.

The emergence of AI directly propels this work into a "breakthrough moment".

Through AI models, it is possible to train on billions of different protein sequences and their underlying structures, completely simulating proteins, eliminating the need for expensive molecular dynamics simulation calculations.

In media interviews, executives from Google DeepMind and NVIDIA have stated that the availability of large amounts of training data, the explosion of computing resources, and the advancement of AI algorithms have collectively sparked the potential of AI in drug development.

Powell said:

"These three elements have come together for the first time, which was impossible five years ago."

This has also sparked investment enthusiasm. According to Pitchbook data, since 2021, there have been 281 venture capital deals for global AI drug development startups, with investments totaling $7.7 billion.

Data volume is a major bottleneck

However, the high computational requirements for the process of fully simulating proteins through AI large models, and the need for sufficient training data, remain a major bottleneck.

Anna Marie Wagner, AI lead at synthetic biology company Ginkgo Bioworks, stated that new basic models like GPT rely on reinforcement learning, a learning process that imitates humans repeatedly training to achieve goals, and relies more on high-quality massive data.

Pushmeet Kohli, Chief Scientist at DeepMind, vividly described the pain point of data volume:

"Garbage in, garbage out."

Furthermore, although the potential of applying AI to drug development is significant, there is still a long way to go before truly entering the pharmaceutical market.

Reportedly, the U.S. Food and Drug Administration (FDA) has approved over 100 clinical trials of drug candidates developed using AI or machine learning, but it may take several years for them to be marketed