Towards the Integration of Artificial Intelligence and Advanced Bioinformatics Tools for Selecting miRNA Molecular Signatures in Type 1 Diabetes

Abstract

Type 1 Diabetes (T1D) is an autoimmune condition characterized by the destruction of pancreatic beta-cells, frequently diagnosed during childhood or adolescence. Recent research has identified microRNAs (miRNAs) as critical molecular regulators involved in the disease’s pathogenesis, suggesting their potential as biomarkers and therapeutic targets. Traditional biological investigations of miRNAs are often costly and time-consuming; thus, this thesis proposes a cost-efficient, bioinformaticsdriven framework as a first-line strategy for narrowing the scope of experimental validation. We integrate advanced in silico methods with machine learning (ML) and deep learning (DL) algorithms to identify and classify T1D-associated miRNA signatures based on gene expression and miRNA expression profiles. This work simulates a multi-omics analytical approach—despite the absence of matched multiomics datasets from the same individuals—by merging heterogeneous data sources to extract biologically meaningful insights. The proposed pipeline ultimately delivers two major outcomes: a biologically interpretable pool of candidate miRNAs, and high-performing classification models capable of clinically distinguishing miRNA signatures relevant to T1D. Our models were validated on both murine and human datasets and demonstrated consistent and optimized performance, supporting the potential of integrative computational pipelines in accelerating translational miRNA biomarker discovery.