AI at the service of longevity: OpenAI explores cellular reprogramming 🧬

Published by Cédric,
Article author: Cédric DEPOND
Source: MIT Technology Review
Other Languages: FR, DE, ES, PT

OpenAI is venturing into the field of longevity science with an artificial intelligence model designed to optimize the production of stem cells. This model, named GPT-4b micro, could transform regenerative medicine by improving the efficiency of proteins responsible for cellular reprogramming.

This initiative marks a significant milestone for OpenAI, as it explores AI applications in biological sciences for the first time. In collaboration with Retro Biosciences, a company specializing in longevity research, the model has already shown promising results in the lab. The goal is to accelerate scientific discoveries to push the boundaries of human aging.

Stem cell science revisited

Stem cells, capable of transforming into any type of tissue, are at the heart of regenerative medicine. Researchers use Yamanaka factors, proteins that enable the reprogramming of adult cells into stem cells. However, this process remains inefficient, with less than 1% of treated cells achieving this state.

The GPT-4b micro model has been trained to propose modifications to these proteins, increasing their efficiency. According to OpenAI, the model's suggestions have improved two of the Yamanaka factors, making them up to 50 times more effective in some preliminary tests.

A strategic collaboration

The project emerged from a collaboration between OpenAI and Retro Biosciences, a startup funded with $180 million by Sam Altman, OpenAI's CEO. Retro Biosciences aims to extend human lifespan by ten years by exploring the mechanisms of aging and developing regenerative therapies.

However, this alliance raises questions about potential conflicts of interest, given Altman's personal investments. OpenAI states that Altman was not directly involved in the project and that the collaboration did not involve any financial transactions.

A specialized model for complex problems

Unlike Google's AlphaFold, which predicts protein structures, GPT-4b micro focuses on reconfiguring their sequences to enhance their functions. The model has been trained on data from various species, enabling it to suggest precise and innovative modifications.

Retro Biosciences researchers tested these suggestions in the lab, achieving concrete results in record time. Joe Betts-Lacroix, CEO of Retro, praised the model's speed and efficiency, highlighting its potential to accelerate scientific discoveries.

Prospects and limitations

Although the results are encouraging, they still need to be validated through scientific publications. The model is not publicly available at this time, and OpenAI has yet to decide whether it will be integrated into its existing products or developed separately.

This breakthrough nevertheless opens new perspectives for longevity research and regenerative medicine. It also illustrates the growing role of AI in solving complex scientific problems, while raising ethical and practical questions about its use.

Going further: What is cellular reprogramming?

Cellular reprogramming is a scientific process that transforms specialized adult cells into pluripotent stem cells. These stem cells have the ability to differentiate into any type of tissue, offering immense potential for regenerative medicine.

This technique relies on the use of specific factors, such as the Yamanaka factors, which activate genes capable of "resetting" cells. Discovered by Shinya Yamanaka, this method earned its creator the Nobel Prize in Medicine in 2012.

However, cellular reprogramming remains a technical challenge. The process is slow, inefficient, and only works on a small proportion of cells. Additionally, it can lead to genetic abnormalities, limiting its clinical use.

Today, artificial intelligence offers new solutions to optimize this process. By analyzing complex biological data, AI can suggest precise modifications to improve the efficiency of reprogramming factors, paving the way for new medical applications.

The role of stem cells in human longevity

Stem cells are cells capable of transforming into any type of tissue, making them a powerful tool for regenerative medicine. They play a key role in repairing damaged tissues and maintaining bodily functions, which can directly influence human longevity.

As we age, the body's ability to regenerate tissues declines, leading to degenerative diseases such as arthritis, heart disease, or dementia. Stem cells, thanks to their differentiation potential, can replace aging or damaged cells, thereby slowing the effects of aging.

Stem cells act in two main ways. First, they can differentiate into specialized cells, such as neurons, muscle cells, or blood cells, to replace those that are failing. Second, they secrete growth factors and anti-inflammatory molecules that stimulate the repair of surrounding tissues.

For example, in the case of heart damage, stem cells can transform into cardiac muscle cells, improving heart function. They can also reduce inflammation, a key factor in aging, by modulating the immune response.

Despite their potential, the use of stem cells to extend human longevity still faces challenges. Their ability to differentiate must be tightly controlled to avoid adverse effects, such as tumor formation. Moreover, their effectiveness decreases with age, limiting their use in older individuals.

Recent advances, such as the optimization of Yamanaka factors by AI, could overcome these obstacles. By improving cellular reprogramming, scientists hope to create more effective and safer stem cells, paving the way for therapies capable of slowing, or even reversing, certain aspects of aging.

Thus, stem cells represent a major promise for extending human longevity, provided their potential is harnessed and the associated technical and ethical challenges are overcome.

GPT-4b micro vs AlphaFold: two AI approaches to biology

GPT-4b micro and AlphaFold are two artificial intelligence models designed to solve complex biological problems, but they operate in very different ways. While AlphaFold focuses on predicting protein structures, GPT-4b micro specializes in reconfiguring their sequences to enhance their functions.

AlphaFold, developed by Google DeepMind, uses neural networks to predict the three-dimensional shape of proteins from their amino acid sequences. This capability has revolutionized structural biology, enabling researchers to understand how proteins interact and function in the body.

GPT-4b micro, on the other hand, is a language model trained on protein sequences and protein interaction data. It does not predict protein structures but proposes modifications to their sequences to optimize their performance. For example, it can suggest changes to the Yamanaka factors to improve their ability to reprogram adult cells into stem cells.

Unlike AlphaFold, which relies on a structural prediction approach, GPT-4b micro uses a method inspired by natural language processing. It analyzes protein sequences as "sentences" and proposes "rewrites" to improve their function. This approach is particularly suited to proteins like the Yamanaka factors, whose structure is flexible and difficult to model.

AlphaFold excels in understanding molecular mechanisms, which is crucial for drug development. GPT-4b micro, meanwhile, opens new possibilities for protein engineering by enabling the design of more effective versions of biological molecules.

However, both models have their limitations. AlphaFold cannot accurately predict complex protein interactions, while GPT-4b micro requires experimental data to validate its suggestions. Together, these tools illustrate how AI can complement traditional approaches in biology, accelerating scientific discoveries and opening new avenues for medicine.
Page generated in 0.138 second(s) - hosted by Contabo
About - Legal Notice - Contact
French version | German version | Spanish version | Portuguese version