Bayesian Independent Component Analysis reconstructs independent modules of gene expression
Listed in
This article is not in any list yet, why not save it to one of your lists.Abstract
Transcriptional regulation—the modulation of gene expression in response to environmental stimuli—is fundamental to cellular function. Identifying groups of co-regulated genes helps elucidate gene functions and characterize how an organism has evolved to respond to various stimuli. In previous works, signal processing algorithms have been applied to characterize the transcriptional regulatory modes, known as iModulons, of bacteria. However, these methods do not quantify uncertainty of the results and are difficult to integrate with different sources of information. In this work, we propose a Bayesian model of Independent Component Analysis that addresses these issues by providing a formal structure to quantify the uncertainty of gene activations and membership of co-regulated genes, achieving state-of-the-art alignment with known regulators. Furthermore, we expand this Bayesian model to explain and integrate first multi-strain and then multi-omics data.
Author summary
Understanding how genes are turned on and off is crucial for deciphering how living organisms respond to their environment. Genes often work together in groups, and identifying these co-regulated groups can reveal their functions and how organisms adapt to changes. Previous methods have used complex mathematical techniques to find these gene groups in bacteria, but they come with limitations: they do not measure how confident we can be in the results and are hard to combine with other types of biological information.
In our study, we introduce a new approach using Bayesian statistics to overcome these challenges. This method not only helps us identify groups of co-regulated genes more accurately but also allows us to quantify our confidence in these findings. Additionally, our approach can easily integrate different kinds of data, such as information from various bacterial strains or other biological processes. This makes our method a powerful tool for exploring gene regulation, with potential applications in understanding diseases, developing new treatments and advancing biotechnology.
