Structure-aware machine learning identifies microRNAs operating as Toll-like receptor 7/8 ligands

Area of Science:

• Immunology
• Molecular Biology
• Bioinformatics

Background:

• MicroRNAs (miRNAs) possess a newly identified function as activation signals for membrane receptors, distinct from their gene regulation roles.
• Toll-like receptors (TLR)7/8 detect single-stranded RNA and initiate immune responses.

Purpose of the Study:

• To develop a machine learning approach, BrainDead, for identifying oligonucleotides acting as ligands for TLR7/8.
• To analyze human miRNAs for their potential to induce TLR7/8 signaling and microglia activation.

Main Methods:

• BrainDead was trained on in vitro murine microglia activation data, considering RNA sequence, structure, and dimerization potential.
• The tool was applied to all known human miRNAs to predict their TLR7/8 ligand activity.
• Predicted activities of selected miRNAs were experimentally validated.

Main Results:

• BrainDead successfully predicted miRNA activity, with high agreement between computational predictions and experimental validation.
• Several high-scoring miRNAs were identified as potent inducers of TLR7/8 signaling and microglia activation.
• A subset of validated miRNAs has known associations with Alzheimer's disease.

Conclusions:

• BrainDead is a robust and powerful tool for predicting miRNA function as TLR ligands.
• The study provides novel insights into the mechanisms of miRNA-mediated immune activation.
• BrainDead offers a generalizable machine learning framework for predicting short RNA functions.