Inferring fungal cis-regulatory networks from genome sequences via unsupervised and interpretable representation...

Area of Science:

• Genomics
• Computational Biology
• Mycology

Background:

• Gene expression is largely controlled by transcription factor (TF) binding to regulatory DNA.
• Predicting gene expression from genome sequences is challenging due to non-functional TF motif matches.
• Functional genomics data for fungal TFs is limited, hindering understanding of their regulatory networks.

Purpose of the Study:

• To develop a sequence-only computational approach for inferring fungal regulatory networks.
• To leverage comparative genomics across multiple fungal genomes to identify conserved regulatory signals.
• To predict gene expression and regulatory connections in fungi without experimental data.

Main Methods:

• Utilized gene orthology as a learning signal to infer TF motif-based representations of regulatory regions.
• Employed comparative genomics to identify conserved motif signals across fungal clades, even without sequence similarity.
• Developed sequence-only models to predict gene expression and regulatory relationships.

Main Results:

• The similarity of conserved motif signals predicted gene expression and regulation more effectively than experimental data.
• Successfully inferred known and novel regulatory connections in diverse fungal species.
• Identified a recombination pathway in *Candida albicans* and mating/RNAi pathways in *Neurospora*.

Conclusions:

• Genome sequence analysis alone can generate testable hypotheses about transcriptional regulation in fungi.
• The developed sequence-only approach is scalable and applicable to diverse fungal clades.
• This method advances the study of fungal regulatory networks where experimental data is scarce.