PNIDB

PNIDB catalogs and annotates protein–nucleic acid interactions extracted from Protein Data Bank (PDB) structures and supplies machine-learning-based protein function predictions to support analysis of transcription, translation, DNA repair, and other protein–nucleic-acid-mediated cellular processes.

Key Features:

• Data Source and Structure: Extracts structural data from the Protein Data Bank (PDB) and uses mmCIF keywords to classify nucleic acid-binding proteins and annotate entries.
• Functional Classification: Categorizes proteins into 27 distinct classes, including transcription factors, immune system components, and structural proteins.
• Predictive Analysis: Trains machine learning models on labeled sequences to predict protein functions, with reported 77.43% prediction accuracy validated by 10-fold cross-validation.
• Entry Annotation: Provides per-entry annotations that summarize protein–nucleic acid interactions derived from PDB structures and mmCIF metadata.

Scientific Applications:

• Transcriptional regulation: Supports research on transcriptional regulation by supplying structural interaction data and functional class labels for nucleic-acid-binding proteins.
• Immune response studies: Enables investigation of immune system components and their nucleic acid interactions through classified and annotated entries.
• Structural biology: Facilitates analysis of protein–nucleic acid complexes using experimentally solved PDB structures and derived annotations.
• Viral infection and drug target research: Aids studies of viral infections and the identification of novel drug targets via function prediction of nucleic-acid-binding proteins.

Methodology:

Integrates structural data from PDB with functional annotations derived from mmCIF keywords; trains machine-learning algorithms on labeled sequence data; validates predictions by 10-fold cross-validation reporting 77.43% accuracy.