NABhClassifier Server: A Tool for the Identification of Helical Nucleic Acid-Binding Sequences in Proteins

This work introduces NABhClassifier, an efficient web server for the identification of short helical segments with nucleic acid-binding potential (NABh) in proteins. Unlike conventional predictors that evaluate entire proteins, NABhClassifier specifically targets the detection of binding domains, which are often the critical determinants of protein-nucleic acid interactions. The prediction pipeline is fully automated and begins with the extraction of helical segments (≥ 6 residues enriched in characteristic amino acids) from protein sequences. For each helix, 20 molecular descriptors are calculated, capturing physicochemical and sequence-based properties. Among these, the frequency of basic amino acids, redox potential, and isoelectric point emerged as the most influential features for classification. The predictive framework integrates eight high-performance machine learning models, each benchmarked for high accuracy. Their outputs are combined into the NABh Index (I_NABh), a consensus score that aggregates results across all models and provides graduated confidence levels for classification. The server provides a user-friendly interface and returns predictions within seconds per protein sequence, enabling rapid large-scale analyses. Its performance was rigorously validated using curated datasets of DNA and single- and double-stranded RNA-binding proteins, as well as genome-wide datasets from multiple species. The tool achieved high recovery rates, including 95-96% for RNA-binding proteins and 93-99% for nucleic acid-binding proteins in genomic data. Overall, this tool is a robust, accurate, and freely accessible resource that supports the discovery of novel biotechnological and biomedical applications. In addition, by enabling the precise identification of nucleic acid-binding helices, the tool offers valuable insights into protein function, molecular interactions, and evolutionary conservation, representing a significant advance in computational protein annotation. The NABhClassifier server, along with detailed tutorials, is available at: http://143.54.25.149.