Multi-scale haplotype signatures distinguish among evolutionary processes
The paper introduces SHADES, a robust deep learning framework that leverages multi-scale haplotype signatures to accurately distinguish among various evolutionary processes, outperforming existing methods and successfully identifying both known and novel adaptive signals in European human genomes.
Original paper licensed under CC BY 4.0 (https://creativecommons.org/licenses/by/4.0/). This is an AI-generated explanation of a preprint that has not been peer-reviewed. It is not medical advice. Do not make health decisions based on this content. Read full disclaimer
The story of how humans became who they are is written in the DNA inside our cells. This genetic code is not a static instruction manual; it is a living record of history, constantly being edited by the forces of nature. Sometimes, a random change in the DNA gives an individual a survival advantage, allowing them to live longer and have more children. Over many generations, this helpful change spreads through the population, leaving a distinct mark on the genetic landscape. Scientists call this process natural selection. However, the genome is also shaped by other forces, such as the random shuffling of genes when parents pass them to children, or the mixing of populations when groups migrate and interbreed. Distinguishing between these different forces is like trying to tell the difference between a fingerprint left by a specific person and a smudge made by a random hand. For a long time, it has been difficult to tell these marks apart, especially when they overlap or when the history of a population is complex.
A team of researchers at Florida Atlantic University has developed a new way to read these genetic stories with much greater clarity. They created a computer program called SHADES, which uses a type of artificial intelligence known as deep learning to look at patterns in human DNA. Instead of just counting how many times a gene appears, the program looks at the arrangement of genetic variants across large stretches of DNA, much like a detective examining the layout of a crime scene rather than just counting the number of people present. The researchers trained this system using millions of simulated genetic histories, teaching it to recognize the unique signatures left by four different evolutionary processes: neutral changes that happen by chance, rapid sweeps where a beneficial gene takes over, balancing selection where multiple versions of a gene are kept in the population, and adaptive introgression, which occurs when humans acquire helpful genes from other ancient human groups.
The results of this training were striking. When tested against other existing methods, the new system proved to be significantly more accurate at identifying which evolutionary force was at work. It could distinguish between a gene that was recently swept through a population and one that was introduced from an outside group, a task that had previously confused many other tools. The system remained reliable even when the data was incomplete or when the rates of genetic change varied across the genome, conditions that often cause other methods to fail. This robustness suggests that the program has learned the true underlying patterns of evolution rather than just memorizing specific examples.
To see if this tool worked on real-world data, the researchers applied it to the genomes of people of European descent from the 1000 Genomes Project. The system immediately confirmed several well-known stories of human adaptation. It correctly identified the region of DNA responsible for lactase persistence, the ability to digest milk sugar into adulthood, which is a classic example of a gene sweeping through a population after the domestication of dairy animals. It also found the genetic signatures of balancing selection in the major histocompatibility complex, a group of genes crucial for the immune system that has maintained high diversity for thousands of years to fight off a wide variety of pathogens. Furthermore, it detected signals of adaptive introgression in genes related to skin pigmentation, confirming that humans acquired some of their physical traits by interbreeding with ancient relatives like Neanderthals.
Beyond these famous examples, the program uncovered a wealth of new candidates for human adaptation that had not been previously highlighted. It pointed to genes involved in immune defense, inflammation, and the maintenance of the skin barrier, suggesting that our ongoing battle with pathogens and our need to adapt to different environments have been constant drivers of our evolution. For instance, the system highlighted genes that help the body respond to viral infections and others that regulate how the skin protects us from the outside world. These findings provide a clearer picture of how natural selection has shaped the human genome, revealing a complex interplay of forces that have helped us survive and thrive in diverse environments.
The success of this approach lies in its ability to look at genetic data at multiple scales simultaneously. By examining small windows of DNA to catch fine details and larger windows to see the broader picture, the program can separate subtle signals that would otherwise be lost in the noise. This multiscale view allows it to tell the difference between a gene that is being actively selected for and one that is simply changing due to random chance or the history of population movements. The researchers found that their method was particularly good at distinguishing between different types of selection, a task that had proven difficult for previous tools.
This work represents a significant step forward in our ability to decode the evolutionary history written in our DNA. By combining advanced computer science with the study of human genetics, the researchers have created a tool that can sift through the vast complexity of the genome to find the specific stories of adaptation. The findings not only confirm what we already knew about human evolution but also open the door to discovering new chapters in our history, particularly those involving our immune system and our interaction with the environment. As the researchers make their software available to the scientific community, it promises to accelerate the discovery of how natural selection has shaped not just humans, but potentially many other species as well.
Drowning in papers in your field?
Get daily digests of the most novel papers matching your research keywords — with technical summaries, in your language.