A Bayesian Multi-Species Approach Infers Gene Regulatory Networks Across Non-Model Organisms

Gene expression regulation by transcription factors is essential for cells to maintain balance amidst environmental cues. Models that predict these regulatory interactions between TFs and controlled genes can shed light on these intricate processes. Such models are invaluable assets, providing testable hypotheses of regulatory interactions, TF function, and expediting the study of uncharacterized TFs.

Nonetheless, the computational complexity of inferring these models arises from the extensive data required due to the myriad potential states of the regulatory network. Bacterial genomes contain hundreds of transcription factors, with many interactions necessitating extensive functional genomic datasets for accurate inference. The situation becomes even more challenging for understudied organisms, species that would greatly benefit from an inferred network for biological discoveries, where the scarcity of available data severely impedes effective inference.

To tackle this issue, researchers have developed GRN-BMuSeR, a groundbreaking multitask approach to gene regulatory network inference that utilizes gene orthology between closely related species to enhance inference performance. The researchers assessed its effectiveness using a dataset from Bacillus subtilis, a well-studied bacterium, and noted improved performance in multitask scenarios.

The model was then applied to simulated data, demonstrating its potential in multi-species contexts. Applying the model to infer gene regulatory networks (GRNs) and explore predictions for two hypersaline-adapted archaeal species, Halobacterium salinarum and Haloferax volcanii, proved fruitful. The rich dataset from Halobacterium salinarum was used to inform the GRN inference of Haloferax volcanii, which had a more limited genomic dataset available.

By generating a vast compendium of gene expression data for Hfx.volcanii for GRN inference input, the researchers discovered hundreds of novel TF functional predictions. The resultant network predictions were found to be in agreement with known TF functions, and the researchers posit that their findings serve as a framework to generate testable hypotheses that will guide experimental work and accelerate discovery in these understudied species.