Deep learning inference of miRNA expression from bulk and single-cell mRNA expression.

Area of Science:

• Genomics
• Bioinformatics
• Computational Biology

Background:

• Studying microRNA (miRNA) activity at the single-cell level is challenging due to technological limitations in capturing miRNAs.
• Existing methods struggle to accurately profile miRNA expression in individual cells.

Purpose of the Study:

• To develop novel deep learning models for predicting miRNA expression from single-cell mRNA data.
• To assess the performance of these models against current state-of-the-art approaches.

Main Methods:

• Introduction of two deep learning models: Cross-modality (CM) and single-modality (SM), utilizing encoder-decoder architectures.
• Prediction of miRNA expression at both bulk and single-cell levels using messenger RNA (mRNA) data.
• Evaluation against the miRSCAPE approach using bulk and single-cell datasets.

Main Results:

• Both CM and SM models demonstrated superior accuracy compared to the miRSCAPE method.
• Incorporating miRNA target information significantly boosted model performance.
• Models utilizing all genes showed lower performance than those incorporating target information.

Conclusions:

• CM and SM models are effective tools for predicting miRNA expression from single-cell mRNA data.
• These models overcome current limitations in single-cell miRNA analysis.
• The findings pave the way for more comprehensive single-cell miRNA expression studies.