Inferring microRNA regulation: A proteome perspective.

Area of Science:

• Molecular Biology
• Genomics
• Bioinformatics

Background:

• Post-transcriptional gene regulation in multicellular organisms heavily involves microRNAs (miRNAs).
• Current methods for predicting miRNA-regulated genes, primarily focusing on 3 -untranslated region (UTR) features and seed matches, yield inconsistent and often inaccurate results.
• Understanding the principles governing miRNA gene regulation remains a significant challenge in molecular biology.

Purpose of the Study:

• To develop a novel computational framework for predicting miRNA-regulated genes based on protein-encoded information.
• To improve the accuracy and cross-species generalizability of miRNA target prediction.
• To identify new miRNA-gene interactions and explore patterns of miRNA regulation across different protein families and gene duplication events.

Main Methods:

• An automated machine learning framework was developed, utilizing protein sequence and diverse functional annotations.
• Models were trained on experimentally validated data across multiple organisms.
• Tens of millions of features were extracted and ranked from various data modalities to identify key predictive indicators.

Main Results:

• The protein-centric model demonstrated high predictive performance within individual organisms and generalized effectively across species.
• The proposed protein model outperformed traditional genomic models, with a unified model showing further improvement.
• Specific protein families, like G-protein coupled receptors (GPCRs), were identified as largely unregulated by miRNAs, contrasting with membranous and disease-related proteins.
• Evolutionary conservation among paralogs did not correlate with coordinated miRNA regulation, indicating functional divergence in regulation tendencies.

Conclusions:

• Protein function is a highly informative feature for predicting post-transcriptional miRNA regulation, offering a new paradigm beyond RNA-based approaches.
• The developed machine learning framework provides a robust and accurate method for identifying miRNA-regulated genes across diverse species.
• Understanding miRNA regulation based on protein characteristics offers new insights into gene expression control and evolutionary dynamics of gene families.