RaptorX

RaptorX predicts three-dimensional protein structures and structural properties, including 8-class secondary structures, solvent accessibility, secondary structures of RNA, and disordered regions, to provide accurate structural information for sequences lacking close homologs in the Protein Data Bank (PDB).

Key Features:

• 8-Class Secondary Structure Prediction: Predicts eight-class secondary structure using a probabilistic method based on conditional neural fields (CNFs).
• Advanced Probabilistic Modeling: Employs CNFs to capture relationships between sequence features and secondary structure and the interdependencies among adjacent residues.
• Integration of Non-Evolutionary Information: Incorporates non-evolutionary information in addition to sequence profiles to improve prediction accuracy.
• Solvent Accessibility and RNA Secondary Structure: Predicts solvent accessibility and the secondary structures of RNA in addition to protein secondary structure.
• Disordered Region Prediction: Predicts disordered regions within protein sequences.
• Performance on Low-Homology Sequences: Designed to handle protein sequences that lack close homologs in the PDB.

Scientific Applications:

• Drug Discovery: Provides structural information to support target characterization and structure-based drug design.
• Functional Annotation: Assists in annotating protein function by supplying predicted secondary and tertiary structural features.
• Disease Mechanism Studies: Aids investigation of structure–function relationships relevant to disease mechanisms.
• Novel Protein Analysis: Enables structural analysis of novel proteins and sequences with few or no homologs.

Methodology:

Uses conditional neural fields (CNFs), a probabilistic graphical model, to model dependencies within protein sequences and predict 8-class secondary structure; integrates sequence profiles and non-evolutionary information; evaluated on benchmark datasets CB513 and RS126 where it outperformed SSpro8.