Who Will Control the Rights to Life-Saving AI-Generated Proteins?

 

By James Myers

The two-day global Artificial Intelligence Action Summit hosted by France earlier this month underscored the essential role that AI is now playing in critical functions like health care. 

Including discussions on international governance of AI and the future of online and AI safety, the summit also highlighted growing concerns about the unequal sharing of AI’s benefits and the need for guardrails to prevent abuses of the technology’s power. The backlash against regulation by some global leaders attending the summit, the July 2024 antitrust judgment against Google for operating an illegal monopoly, and the recent actions of some technology leaders in response to the administration that gained power last month in the U.S., heighten these concerns.

Particularly powerful and potentially extremely profitable AI applications are in medicine, where machine learning has given AI the ability to predict and design proteins that are essential to the proper functioning of the human body and other living organisms. It’s a power that holds a real promise of curing debilitating autoimmune conditions and deadly diseases like cancer.

 

 

With its eye on profits, Alphabet, which owns Google, Chrome, YouTube, and Android, is betting heavily on its DeepMind division that makes the protein-predicting application called AlphaFold. Employing machine learning and innovative algorithms, AlphaFold has succeeded in predicting the structures of over 200 million proteins, which is practically all known proteins. Additionally, DeepMind’s technology has been used to create 2.2 million new crystal structures, an important development in materials science.

 

 

Alphabet, which made a record-smashing profit of $100 billion in 2024, earned 75% of its $350 billion revenue last year from advertising made possible by data freely taken from people around the world who use its products. The company is investing some of those profits in advancing the capabilities of AlphaFold to generate further breakthroughs in biology. 

Rapid advancements in chemistry, machine learning, and biology are increasing the probability that AlphaFold or other companies with powerful technologies might soon conquer diseases like cancer. Cancer is the second-leading cause of death globally, and in 2021 claimed almost 10 million lives in spite of decades of research at the cost of hundreds of billions of dollars. Manipulating the structure of proteins to cure a disease like cancer would be an incredible breakthrough in medicine that could generate astronomical profits for a company that holds the legal patent on the particular protein or proteins that could save millions of lives.

The battle against cancer has already advanced significantly, from treatments such as immune therapy that have recently become available. Immune therapy coordinates the immune system in battling cancerous cells, slowing or stopping their progression. 

The awarding of half of the 2024 Nobel Prize in Chemistry to Google DeepMind CEO Demis Hassabis and director John Jumper highlights the importance of the company’s achievements in predicting protein folding. The other half of the 2024 Chemistry prize was awarded to David Baker, who developed algorithms that generated the first artificially-engineered protein in 2003. Also in 2024, the Nobel Prize in Physics was awarded to John Hopfield and Geoffrey Hinton for using “tools from physics to develop methods that are the foundation of today’s powerful machine learning.”

To engineer a protein, Baker used a process called “RF diffusion” to add random data, or noise, to a known protein structure.

AI trained with machine learning was then applied to the noise to calculate a new protein configuration. The process has been used to synthesize safer anti-venom treatments for snake bite victims, and its potential application to cancer and autoimmune diseases are now being investigated.

 

What rights will laws provide to commercial interests in life-saving protein treatments?

While DNA, RNA, and gene sequences that operate the genome of humans and other living species cannot be patented, in many jurisdictions patent protection is available for technologically-developed proteins with specific applications, as well as the methods for their application. There are, however, some important differences between the laws of some jurisdictions which could lead to disputes over rights to novel protein-folding technologies.

Commercial interests are already organizing to protect their intellectual property rights over protein design. For example, Protein Industries Canada was established in 2018 as an organization of Canadian companies to foster development of plant proteins and advance legal protection of their processes.

How would a company like Alphabet act, if it held a patent for a cure to cancer? 

Alphabet has provided some clues to how such a future might unfold. While AlphaFold’s protein folding code was publicly available for all researchers in its first and second versions, the company introduced restrictions with the latest version, AlphaFold3. 

While anyone with a Google account can use AlphaFold3 without charge for non-commercial applications, DeepMind retains the rights to commercial uses. As Nature noted in a May 2024 editorial, DeepMind did not publish the code with its release of AlphaFold3, providing instead a limited description of it. 

Nature made a controversial decision to publish the release of AlphaFold3 without the underlying code being made available for peer review. In breaking with its normal publishing practices, Nature stated, “This is an opportunity for an important conversation among all research stakeholders at a time when the majority of global research is privately funded.” 

Nature noted that while DeepMind’s AlphaFold2 had been developed with the assistance of the publicly-funded European Molecular Biology Laboratory’s European Bioinformatics Institute, AlphaFold3 was created in collaboration with Isomorphic Labs, a drug-development company owned by Alphabet. Further, DeepMind placed daily limits on the number of protein-folding predictions that public users can generate. 

 

 

Where is protein-folding prediction and technology heading?

Quantum computers are poised to revolutionize the pharmaceutical industry with their ability to manipulate combinations and permutations of data, like that used in protein folding predictions, far more quickly than today’s fastest supercomputers.

Funded by advertising dollars, Alphabet is one of the world’s leading developers of quantum computing technology. As we noted in our article, Giant Steps Have Been Taken Toward Our Quantum Computing Future, Google Quantum AI made global headlines last December when it announced that its quantum chip, named Willow, had succeeded in performing a calculation in five minutes that would have required ten septillion years on a supercomputer. 

Quantum computing could amplify many times over the already powerful application of AI in protein folding prediction.

 

 

While current quantum computer prototypes continue to suffer limitations from their inability to maintain stable circuits, recent breakthroughs suggest that a fully-functioning quantum computer might be available within five years, or even sooner. While the world awaits the power of the quantum computer, AI has already delivered remarkable capability to predict patterns and variations in protein folding. 

Proteins are comprised of amino acids, which are organic compounds that combine in peptide bonds to form strings that fold in different shapes to make proteins with specific functions. From a total of close to 500 amino acids, there are 22 in the genetic code of living organisms, with a massive number of different possible configurations in folding.  

For example, chemical and electrical reactions in a string of 35 amino acids can result in an astronomical 10³³ different ways of folding, an impossibly long task for today’s fastest supercomputers to predict. However, it’s the folded shape of the protein that determines its specific function, such as that of hemoglobin which carries oxygen in the blood.

The winner of the 1962 Nobel Prize in Chemistry, John Kendrew, was the first scientist to determine the structure of protein long before the advent of AI. It took Kendrew twelve years to determine the structure of one protein, by reconstructing its shape from the diffraction pattern of x-rays.

AlphaFold’s innovation was in the use of machine learning to decipher patterns in pairs and triplets of amino acids. The technology represents connections of amino acids by measuring the distances, twisting, and torsion of their combinations to determine the geometry of the shape and function of the proteins they create.

Who will control life-saving engineered proteins?

Science, technology, and AI are promising to enable the engineering of proteins that could restore biological functions and save lives by controlling or eradicating diseases. 

Further investment will be required to advance the processes and medical applications, and developers will naturally want financial compensation for their efforts. The question is, how much compensation is fair and who will control the technology that could save many lives?

There are many examples of life-saving medications and treatments that are not available to many because the profit motive has kept some expensive life-saving medications from those in need, especially in less wealthy parts of the globe. 

For instance, while HIV is now a generally well-controlled and often preventable chronic condition in wealthy nations, where it rarely advances to deadly AIDS, the predicament in Africa is dire. As of 2023, there were over 25 million Africans living with HIV, accounting for two-thirds of the global total in an epidemic that has infected more than 10% of adults in some particularly hard-hit countries. In sub-Saharan Africa, uncontrolled HIV often leads to AIDS, which claimed the lives of 1.2 million people in 2011 out of a global total of 1.7 million AIDS deaths that year.

According to the United Nations, in 2022 AIDS claimed one life every minute, and in areas like north Africa only about half of people infected with HIV were receiving treatment that would prevent the onset of AIDS.

In 2015, so-called “pharma-bro” Martin Shkreli provided a stark example of how the pursuit of profit can come at the expense of life-saving treatments. That year, his company Turing Pharmaceuticals acquired the manufacturing licence for the antimalarial and antiparasitic drug Daraprim, which had been available for many years. Shkreli raised Daraprim’s price by more than 5,500%, from $13.50 to $750.00 per tablet. In defending the unprecedented price increase, Shkreli stated, “If there was a company that was selling an Aston Martin at the price of a bicycle, and we buy that company and we ask to charge Toyota prices, I don’t think that that should be a crime.”

In raising the question of global AI governance, the Artificial Intelligence Summit earlier this month highlighted a crucial issue as AI plays an increasingly pivotal role in human health.

With life-saving proteins, who will hold the key to human health and at what price?

Craving more information? Check out these recommended TQR articles:

• Deepfakes: The Technology of Deception and Quantum’s Uncertain Future With the Problem
• Is Data-Intensive AI Facing a Memory Limit? New Approaches Might Provide Solutions
• Swarms and Collective Intelligence: Bridging Biology, Sociology, and Computer Science
• Adaptive Intelligence: The Power of Genetic Mutations and Neural Networks
• What Comes Next, After Profit-Driven OpenAI Cracks the Intelligence Code?
• In Praise of Human Intelligence For All Time
• Trust is Fragile: Google’s Antitrust Loss and the Global Windows Failure in July are Warnings to AI Investors About the Future Value of Trust