sChemNET: a deep learning framework for predicting small molecules targeting microRNA function.

Area of Science:

• Biochemistry
• Computational Biology
• Genetics

Background:

• MicroRNAs (miRNAs) are crucial regulators implicated in various human diseases, including cancers and infectious diseases.
• Modulating miRNA activity or their target genes with small molecules presents a therapeutic avenue for disease treatment.
• Predicting small molecule-miRNA interactions is challenging due to limited datasets.

Purpose of the Study:

• To develop a generalized deep learning framework, sChemNET, for predicting small molecules that affect miRNA bioactivity.
• To overcome data sparsity by enabling the model to learn from extensive chemical structure information.
• To experimentally validate the framework's predictions in biological systems.

Main Methods:

• Developed sChemNET, a deep learning framework utilizing chemical structure and sequence information.
• Employed an objective function to learn chemical space from large, unlabeled chemical datasets.
• Conducted experimental validation of predicted small molecules targeting miR-451 and the miR-181 network in zebrafish and in-vitro models.

Main Results:

• sChemNET successfully predicted small molecules affecting miRNA bioactivity.
• Experimental validation confirmed the efficacy of predicted small molecules in biological assays.
• Demonstrated the framework's capability to identify bioactive small molecules for miRNA targeting.

Conclusions:

• The developed sChemNET framework offers a powerful approach for identifying small molecules that modulate miRNA activity.
• This tool can accelerate the discovery of novel therapeutics for miRNA-related disorders.
• The machine-learning framework shows promise for applications across human and mammalian organisms.