PSIPRED

PSIPRED predicts protein secondary structure, transmembrane topology, and fold recognition to support structural and functional analysis of protein sequences.

Key Features:

• Secondary Structure Prediction: Employs a two-stage neural network that uses position-specific scoring matrices generated by PSI-BLAST and reported Q3 scores of 76.5%–78.3% in CASP3.
• Transmembrane Topology Prediction: Implements MEMSAT-2 for predicting transmembrane topology of membrane proteins.
• Protein Fold Recognition: Incorporates GenTHREADER, a sequence profile-based fold recognition method for predicting three-dimensional folds by comparison to known structures.
• Scalability and Integration: Provides SOAP and XML-RPC web server access and HADOOP packages for scalable execution of prediction methods.

Scientific Applications:

• Structural Biology: Provides secondary-structure and topology predictions to inform interpretation and modeling of protein structures.
• Functional Annotation and Evolutionary Studies: Supports functional annotation and evolutionary analysis through fold recognition and sequence-profile comparisons.
• Drug Discovery: Supplies structural hypotheses for target proteins that can inform the design of therapeutic agents.

Methodology:

Secondary structure prediction uses a two-stage neural network trained on position-specific scoring matrices derived from PSI-BLAST; validation included CASP3 (Q3 76.5%–78.3%) and subsequent testing sets comprising 187 unique folds; fold recognition uses sequence profile-based methods (GenTHREADER); transmembrane topology prediction uses MEMSAT-2; scalable deployment uses SOAP, XML-RPC, and HADOOP.