Inflammation-Linked Aging Signals In Frozen Single-Cell Foundation Models: Donor-Aware Detection And Robustness Testing

Researchers developed a new method to determine if aging signals found in advanced AI models, which were not specifically trained on age, represent true biological processes or are simply artifacts from how data was collected.

Scientists are using powerful AI models, known as foundation models, to understand complex biological data, like the information from individual cells. These models are trained on vast amounts of data but aren’t explicitly taught about things like a person’s age. A new study introduces a crucial framework to test whether these models can still pick up on real biological signals related to aging, or if what they detect is just a trick of the data.

Using this new testing approach, researchers found that these AI models do indeed contain hidden information about aging, even though they were never directly shown chronological age during their training. This aging signature is particularly linked to inflammation pathways in the body, specifically those involving NF-κB and IFN-γ, which are already known to be important in the aging process.

While these AI models don’t necessarily predict age better than simpler statistical methods, they offer a unique way to interpret the underlying biological mechanisms of aging. The new framework includes rigorous checks, such as carefully re-sampling cells to ensure that any observed aging signals are truly biological and not just due to differences in cell types between younger and older individuals. This work helps us better understand how AI can uncover complex biological insights and provides a robust method for validating these discoveries.

---

Source: link to paper