SMTRI: A deep learning-based web service for predicting small molecules that target miRNA-mRNA interactions

Area of Science:

• Biochemistry
• Computational Biology
• Pharmacology

Background:

• Mature microRNAs (miRNAs) are crucial regulators of gene expression, implicated in various diseases.
• Targeting miRNAs with small molecules offers a promising therapeutic strategy.
• Current drug screening methods are often time-consuming and labor-intensive.

Purpose of the Study:

• To develop an efficient computational program for *in silico* drug discovery targeting mature microRNAs (miRNAs) and their mRNA interactions.
• To identify small molecules that specifically target the structural motifs of miRNA-mRNA complexes.

Main Methods:

• Utilized a convolutional neural network (CNN)-based approach named SMTRI.
• Trained SMTRI to predict small molecules targeting RNA secondary structural motifs formed by miRNA-mRNA interactions.
• Validated SMTRI's performance on three independent testing sets.

Main Results:

• SMTRI demonstrated superior performance compared to state-of-the-art algorithms, with AUC improvements of 12.9%-30.3% and accuracy increases of 2.0%-18.4%.
• Case studies involving four published, experimentally validated RNA-targeted small molecules confirmed SMTRI's reliability.
• The study highlights the potential of targeting unique functional loops in miRNA-mRNA interactions.

Conclusions:

• SMTRI provides an efficient and reliable computational tool for accelerating the discovery of novel RNA-targeted drugs.
• This *in silico* approach can significantly reduce the time and cost associated with traditional drug screening.
• The findings support the development of targeted therapies for miRNA-related diseases by precisely targeting specific miRNA-mRNA interactions.