Medicine and artificial intelligence are fields that seem to have a rather jarring intersection. Computer scientists obviously don’t have medical degrees, but have been able to make incredible strides in improving disease detection with an unforeseen frontrunner protein structure prediction in biology model: AlphaFold that has just made an immense leap forward.
AlphaFold– 3 is a novel A.I. algorithm that can predict the structures and interactions of all of life’s molecules, ranging from proteins to nucleic acids. Developed by Google’s subsidiary, Google DeepMind, and Isomorphic Labs, this software has allowed scientists to reach newfound heights in drug discovery and medical research.
The model was first released in Nature, a prestigious scientific journal known for its reputability and transparency standards. However, AlphaFold– 3’s publication on May 8, in May 2024, however, has sparked significant controversy on corporate-washed scientific research and what it means for a study to be peer-reviewed.
As the third iteration of AlphaFold, the model’s scope is much broader than its older counterparts. It can now predict the structure of all of life’s molecules, including proteins, lipids, carbohydrates, and nucleic acids. In layman’s terms, this algorithm can accurately speculate aid in determining how certain diseases impact the body and, by extension, determine how various types of experimental medicine treatment will functions in the body. This success is partly attributed to the immense training initiative in which scientists fed the algorithm the structure of 100,000 proteins.
“It’s our gift to humanity and a demonstration of the benefits A.I. can bring to society,” Google DeepMind and Isomorphic Labs CEO Dennis Hassabis posted on X.
The original AlphaFold model was first released at the 13th Critical Assessment of Techniques for Protein Structure Prediction event, which began on May 1, in 2018. A few years later in 2021, AlphaFold– 2 was released, with a more accurate machine– learning model for even larger protein systems, was released on July 22, 2021.
AlphaFold has been recognized as one of the most influential AI models today, with computational biologists behind AlphaFold accepting the 2022 $3 million Breakthrough Prize, the 2023 Albert Lasker Basic Medical Research Award, and first place at CASP14. Moreover, the three versions predicted over 200 million protein structures in compilation, aiding over 20,000 studies.
“The capabilities of A.I. never cease to fascinate me,” rising junior Jonah Misaghi said. “It’s truly amazing what scientists can do in this day and age, and I’m excited to see the clinical applications of this technology.”
Typically, machine learning model publications are required to include the corresponding code to enable result replication and verification. However, the published study on AlphaFold 3 only provided pseudocode, a logical description of steps that break down an algorithm’s inner workings –that is, no actual code. Many criticize Nature for its permission of such a deliberate lack of method transparency and yielding to corporate interests. The journal responded by citing the biosecurity risks of malicious groups initiating biological attacks with this new technology.
The general consensus in the field, however, is that Google DeepMind’s decision to keep its algorithm from the public is a major disservice to the scientific community, and is a primary example of greed taking precedence over discovery –an idea that many previously thought academia was immune from.
“AlphaFold 3 is a monumental achievement, yet its restricted access highlights a growing tension between innovation and transparency,” rising senior Ethan Shirazi said. “Progress must be paired with public trust.”
Nonetheless, the technology is available to the general public through a user-friendly website that requires no coding experience whatsoever. Although it is widely accessible, the technology is not available for commercial use, limiting users to 10 predictions per day and not allowing structure predictions of custom proteins. In essence, researchers are only able to use the model for small jobs. ESMFold, AlphaFold’s top competitor, is the most commonly cited alternative; Although it is an open-source substitute, its abilities are limited to proteins, and cannot predict the structures of other biomolecules.
A.I. has clearly shown to be a powerful tool in just about any field. Debates regarding biosecurity and the consequences of propagating data that can be used maliciously have plagued disclaimers and legal bounds of the use of such volatile models, and will continue to pose a major problem for researchers.
When it comes to Nature’s negligence of the scientific process and Google DeepMind’s confidentiality, the capability of the model speaks for itself. The fact of the matter is that AlphaFold 3 is a product with a lot of profit potential; giving away the ingenuity behind its invention for free is simply irrational from a corporate standpoint, even if that slows scientific progress momentarily.
“You can’t blame these companies for attempting to maximize profit,” rising junior Eliav Sehati said. “If we forced everyone to give away their research for free, then from a monetary perspective, few people would be motivated to publish research.”
Most computational biology research endeavors, among other fields, are funded by private companies, and surrendering promising products such as AlphaFold 3 could cause a major decline for the conglomerates sponsoring this groundbreaking research.
Without a doubt, humanity’s grasp on medical innovation is improving. A.I. is just one of many tools which can help us fight disease and learn more about the underpinnings of the microscopic factories that keep our bodies working. Despite the controversy, AlphaFold 3 is an extraordinary development in scientific research that truly has the potential to help us live longer, healthier lives –something every person, scientist or not, can appreciate.
