Year: 2025
Type: article
Abstract: Studies have shown that only about 2% of the genome encodes proteins, while the remaining 98% consists of non-coding RNAs (ncRNAs). Based on length, ncRNAs are classified as small (<200 nt) or long (>200 nt) and play key roles in biological processes. Experimentally verified associations between ncRNAs (miRNAs, lncRNAs, circRNAs) and diseases remain limited, since laboratory studies are costly and time-consuming. Thus, computational approaches have become essential for predicting disease related ncRNAs. Similarly, drug-target interactions are vital for drug discovery, as drugs act by binding to and inhibiting target molecules. Yet, experimental identification of these interactions is expensive, driving the development of computational prediction methods. Microbes also influence human health, with microbiomes playing essential physiological roles. Identifying disease-related microbes is crucial, but experimental approaches are limited by cost and time. Hence, computational methods are widely employed. Overall, computational strategies can be grouped into score functions, network-based algorithms, multi source biological integration, and machine learning. This review highlights machine learning approaches for predicting ncRNA-disease associations, drug-target interactions, and disease-related microbes. It also summarizes key databases and successful methodologies, serving as a guide for future research in this field.
Authors: TOPRAK, Ahmet, ERYILMAZ DOGAN, Esma
Institution: Selçuk University
Language: English
Subfield: Computational Theory and Mathematics
Field: Computer Science
Domain: Physical Sciences
Open Access status: diamond
OpenAlex ID: W7217295607