Researchers have introduced NeRNA-Web, a user-friendly, web-based platform designed for generating structure-aware negative RNA sequences and conducting comparative evaluations in noncoding RNA classification. Published as open-access software, the newly developed system provides an accessible, unified, and reproducible pipeline that enables investigators to generate, examine, and download negative RNA datasets across multiple noncoding RNA classes for downstream computational workflows.
In computational biology, supervised machine learning models are widely applied to noncoding RNA classification tasks. However, the overall performance and predictive validity of these computational models depend heavily on how negative training datasets are constructed. Frequently employed negative generation strategies, including random shuffling and dinucleotide-preserving shuffling, can introduce systematic biases during model training. Despite these known challenges, researchers previously lacked a standardized and accessible framework to generate and systematically compare negative RNA datasets.
To address this methodological challenge, NeRNA-Web implements the previously developed NeRNA transformation. This computational technique incorporates predicted secondary-structure information into an encoded representation to generate transformed negative RNA sequences. In addition to structure-aware generation, the platform incorporates standard sequence generation strategies, such as random shuffling and dinucleotide-preserving shuffling, within a single environment. By bringing these diverse approaches together into a unified pipeline, the platform facilitates direct methodological comparisons between structure-aware transformations and traditional sequence shuffling techniques.
The web-based platform is structured to support a wide range of major noncoding RNA classes, explicitly accommodating microRNA, long noncoding RNA, transfer RNA, and circular RNA. Alongside negative sequence generation, NeRNA-Web provides built-in visualization tools and convenient data export functionalities to assist downstream research workflows.
NeRNA-Web delivers an accessible and extensible framework intended to foster transparency and reproducibility in negative dataset construction for noncoding RNA studies. The authors emphasize that negative training datasets must be carefully selected, characterized, and evaluated to match the exact requirements of intended downstream noncoding RNA classification tasks.
Subject of Research: Web-based generation and comparative evaluation of negative RNA training datasets
Article Title: NeRNA-Web: A web-based platform for structure-aware negative RNA generation and comparative evaluation in ncRNA classification
Article References: Orhan, M. E., Demirci, Y. M., & Saçar Demirci, M. D. (2026). NeRNA-Web: A web-based platform for structure-aware negative RNA generation and comparative evaluation in ncRNA classification. BMC Bioinformatics. https://doi.org/10.1186/s12859-026-06768-7
Image Credits: AI Generated
DOI: 10.1186/s12859-026-06768-7
Keywords: NeRNA-Web, noncoding RNA, ncRNA classification, machine learning, negative dataset generation, RNA secondary structure, NeRNA transformation, dinucleotide-preserving shuffling, random shuffling, miRNA, lncRNA, tRNA, circRNA, comparative evaluation, model generalizability, training data bias, reproducible bioinformatics, supervised learning
News Source: Teresa Odom. (November 18, 2026). NeRNA-Web Generates Structure-Aware Negative RNA Datasets for Machine Learning Classification. Scienmag.



