Genes Under Surveillance: New AI Shows When Genes Are On Or Off

During the body's development, genes need to be turned on and off at the right time for cells to assume their correct functions. When this fails, diseases can arise. Researchers have created a tool called chronODE, which uses mathematics and artificial intelligence to track how genes are regulated over time. With it, they discovered patterns that show how DNA and its control mechanisms work together, which could help better understand brain development and pave the way for new treatments.

During an organism's development, each cell must "decide" its function, for example, whether to become a brain, muscle, or blood cell. These decisions depend on which genes are active or inactive at any given moment. When genes are turned on or off at the wrong time, cells can lose their normal function and even cause disease.

The control of these processes is largely done by mechanisms called epigenetics, which regulate how DNA is read without changing its sequence. Understanding the speed and rhythm at which these genes are activated or silenced is essential to understanding both the healthy functioning of the body and the emergence of problems.

If it is possible to precisely control these processes, it could also pave the way for new treatments that adjust gene activity in a targeted manner.

To study these rhythms, scientists at Yale University, USA, analyze how gene expression and chromatin structure (the way DNA is organized within cells) change over time.

Observing isolated moments alone is insufficient, as this merely provides "still photographs" of the processes. Tracking genes at different times reveals dynamic patterns, as if watching a movie. However, processing these data is challenging because they contain a lot of "noise" and are highly interconnected, requiring sophisticated computational methods to separate the important signals and understand what is actually happening in the cells.

Epigenetic mechanisms, such as chromatin opening or chemical modifications to proteins called histones, follow basic principles. One of them is cooperativity: when chromatin begins to open, the process tends to reinforce itself, further facilitating the opening. The same is true for histone modifications, which can spread rapidly to neighboring regions.

The other principle is saturation: these processes have natural limits and cannot continue indefinitely, as they depend on the quantity of available molecules. These concepts also help us understand how RNA production, which is the copy made of genes, varies over time, demonstrating that gene expression is governed by rules similar to those of chromatin.

Despite advances in the study of RNA production and degradation rates, we still know little about how epigenetic processes change over time and how they directly influence gene activity.

To address this challenge, scientists have developed a new tool called chronODE, which combines mathematical models and artificial intelligence to analyze data from multiple biological layers over time. This tool can smooth out the "noisy" signals from experiments and identify different patterns of gene activation or repression.

The researchers applied chronODE to data from mouse brain development and observed that most genes (~87%) follow a gradual and limited activation pattern, like a logistic curve. Only a small percentage of genes accelerate rapidly and reach high levels of activity, suggesting that chemical constraints prevent all genes from functioning this way.

They also found that genes activated at different stages of development follow different rhythms, with those most essential for brain formation exhibiting the fastest acceleration.

Distribution of kinetic parameters k (y-axis) and b (x-axis) for activated genes (k > 0) in 41 brain cell types. Cell types were grouped based on the day of appearance during the embryonic period (between 8 and 9, 9 and 10, 10 and 11, 12 and 13, and 14 and 15 embryonic days [E]).

Furthermore, chronODE showed that regulatory elements close to genes are important in deciding whether they are activated or silenced, while more distant elements control the timing and pattern of these changes. To further refine the predictions, the scientists used recurrent neural networks, which learned to predict how changes in chromatin are reflected in changes in gene expression.

Overall, this approach paves the way for a better understanding of how genes and their control mechanisms work in a coordinated manner over time. This can help us understand both the normal development of organisms and the emergence of diseases, as well as offer new possibilities for therapeutic intervention.

READ MORE:

The chronODE framework for modelling multi-omic time series with ordinary differential equations and machine learning

Beatrice Borsari, Mor Frank, Eve S. Wattenberg, Ke Xu, Susanna X. Liu, Xuezhu Yu, and Mark Gerstein 

Nature Communications. 16, 7021 (2025). 

https://doi.org/10.1038/s41467-025-61921-9

Abstract: 

Many genome-wide studies capture isolated moments in cell differentiation or organismal development. Conversely, longitudinal studies provide a more direct way to study these kinetic processes. Here, we present an approach for modeling gene-expression and chromatin kinetics from such studies: chronODE, an interpretable framework based on ordinary differential equations. chronODE incorporates two parameters that capture biophysical constraints governing the initial cooperativity and later saturation in gene expression. These parameters group genes into three major kinetic patterns: accelerators, switchers, and decelerators. Applying chronODE to bulk and single-cell time-series data from mouse brain development reveals that most genes (~87%) follow simple logistic kinetics. Among them, genes with rapid acceleration and high saturation values are rare, highlighting biochemical limitations that prevent cells from attaining both simultaneously. Early- and late-emerging cell types display distinct kinetic patterns, with essential genes ramping up faster. Extending chronODE to chromatin, we find that genes regulated by both enhancer and silencer cis-regulatory elements are enriched in brain-specific functions. Finally, we develop a bidirectional recurrent neural network to predict changes in gene expression from corresponding chromatin changes, successfully capturing the cumulative effect of multiple regulatory elements. Overall, our framework allows investigation of the kinetics of gene regulation in diverse biological systems.