GPS: A Computational Intelligence-Based Tool for the Discovery of G4 Motifs with Their Folding Status from G4-Seq Data

G-quadruplexes (G4s) are unique secondary structures formed in guanine-rich DNA or RNA regions, stabilized by Hoogsteen hydrogen bonds. These structures play critical roles in genomic integrity, transcription regulation and telomere protection. Traditional methods for G4 identification, such as circular dichroism spectroscopy and polymerase stop assays, are limited by low throughput, specificity and experimental constraints. To address these challenges, we developed G4 Prediction Server (GPS), a computational tool for precise G4 motif prediction and folding potential assessment. GPS utilizes high-throughput sequencing data (GEO: GSE110582) from six organisms — Arabidopsis thaliana, Drosophila melanogaster, Caenorhabditis elegans, Homo sapiens, Saccharomyces cerevisiae and Danio rerio. Positive samples included K[Formula: see text] and pyridostatin-stabilized genomic regions, while non-G4 regions served as controls. A Random Forest classifier achieves 97% accuracy and an [Formula: see text]1-score of 0.98 for motif identification. For folding potential, GPS employs Random Forest and XGBoost models, assessing features such as loop count, GG/GC content and minimum free energy, with RMSE values between 3.51 and 4.08. Folding stability is confirmed when two out of four criteria are met. GPS integrates RNAfold for secondary structure analysis and provides detailed G4 properties, including loop features and quartet formation. Validated using human G4 X-ray crystallography data, GPS outperformed existing tools like PENGUINN and G4Hunter, establishing itself as a robust resource for drug discovery, gene regulation and genomic stability studies. This user-friendly platform offers unparalleled accuracy and insights into G4 biology, making it indispensable for high-throughput structural and functional analyses.