Overview

DExTER (Domain Exploration To Explain gene Regulation) is a bioinformatics tool designed to automatically identify genomic regions whose nucleotide composition correlates with gene expression levels.

Unlike traditional approaches focusing on short transcription factor binding sites (6-12 bp), DExTER detects Long Regulatory Elements (LREs) that can span tens to hundreds of nucleotides.

This makes it possible to explore a poorly characterized regulatory mechanism: gene regulation driven by the global composition of genomic regions.

The method has shown strong predictive power, explaining a large fraction of gene expression variability, reaching more than 70% prediction accuracy in Plasmodium falciparum.

Method principle

DExTER combines DNA sequences and gene expression data to identify pairs:

(k-mer motif, gene region)

whose frequency is correlated with expression level.

Main steps

1. Segmentation of regions around genes
2. Iterative search for k-mers correlated with expression
3. Informative variable selection (LASSO / machine learning)
4. Construction of a predictive gene expression model

The final model allows:

• identification of candidate regulatory regions
• explanation of expression variability
• prediction of expression of new genes

Biological applications

DExTER enables the study of regulatory mechanisms not detectable with classical motif discovery methods:

• Genomes with few transcription factors
• Post-transcriptional regulation
• Cell-cycle dependent regulation
• Epigenetic regulation linked to DNA composition

Key observations from the study:

• Highly dynamic regulation in Apicomplexa
• More stable regulation in multicellular organisms
• Distinct roles of upstream (transcriptional) vs downstream (post-transcriptional) regions

Input data

• Nucleotide sequences aligned per gene (e.g. ±2 kb around TSS or start codon)
• Gene expression matrix (RNA-seq or microarray)

Output data

• List of candidate regulatory elements (cLREs)
• Enriched motifs and associated regions
• Variable importance
• Predictive expression model
• Correlation scores between sequence and expression

Typical use cases

• Study of non-canonical gene regulation
• Comparative genomics
• Transcriptome interpretation
• Atypical genomes (parasites, plants, protists)

Associated publication

Menichelli C. et al., 2021
Identification of long regulatory elements in the genome of Plasmodium falciparum and other eukaryotes
PLOS Computational Biology
https://doi.org/10.1371/journal.pcbi.1008909

Source code

https://gite.lirmm.fr/menichelli/DExTER