KAUST AI tool Unify analyzes 700 million years of cell evolution

Identifying Biological Parallels Beyond Gene Sequences

Traditional genomic research often relies on single-cell RNA sequencing to compare cell types across different species. This method builds "cell trees" by mapping gene activity, but it frequently struggles when comparing distant species because the specific genes responsible for certain functions diverge over time. Most conventional tools require one-to-one gene matches, which can cause researchers to overlook functional similarities between organisms like humans and mice or fish and flies.

Unify addresses this limitation by moving beyond simple dictionary-style word-for-word translation. Instead, it uses AI models to analyze protein sequences and scientific descriptions of gene functions. As explained by Huawen Zhong, lead author and computational biologist at KAUST, the tool groups genes with similar functions into units called "macrogenes."

“Unify works with AI models that analyze protein sequences and scientific descriptions of gene functions. Genes with certain similarities are grouped into units called ‘macrogenes.’ This allows Unify to recognize cells performing similar jobs, even when their individual genes no longer match.” — Huawen Zhong, KAUST

Reconstructing Evolutionary Relationships Across 125 Cell Types

In a study published in Nature Communications, the KAUST team demonstrated the tool’s capability by reconstructing relationships among 125 cell types from seven different species. The AI successfully distinguished between genes that remained identical, those that evolved to perform new tasks, and genes that emerged independently to fulfill the same biological function.

The model showed particular utility in analyzing immune cells. It identified shared defense tactics across species that would have been invisible to standard comparison methods. In a specific experiment, the team used Unify to predict how human blood cells would respond to a particular immune-signaling protein, using data from mouse lymph-node immune cells as a baseline. The researchers reported that Unify’s predictions were more accurate than those generated by existing computational methods.

Enhancing Human Health Research via Animal Models

The primary application for Unify lies in the clinical translation of laboratory findings. Because human health research frequently depends on data from model organisms, identifying which biological processes are truly conserved is essential for successful experimentation. According to Manuel Aranda, a professor of marine science at KAUST, the tool provides a clearer roadmap for determining the relevance of animal studies to human health and disease.

“A lot of what we know about human biology comes from studying animals like mice, but it is not always clear which findings carry over. Unify helps us identify which discoveries in model organisms are most likely to be relevant to humans, so research can be focused where it is most useful for understanding human health and disease.” — Manuel Aranda, KAUST

Future Expansions for Genomic Mapping

Following the initial success of the model, the KAUST researchers are working to incorporate additional biological layers into the Unify framework. Future updates are expected to include data regarding gene regulation and the physical positioning of cells within tissues. By integrating these factors, the team aims to provide a more comprehensive view of the biological principles shared across life, further refining the ability to map evolutionary trajectories and their implications for human health.