RAPTOR Online

Overview

All the user information is kept confidentally. If you want to submit sequences or check the prediction results, please login in first.

RAPTOR is a successful protein three-dimensional structure prediction program. RAPTOR's performance has been demonstrated in the CASP5 and CASP6 competitions and by BSI's various customers.

Threading Methods

The RAPTOR software suite contains three threading methods:

• NpCore: Each template is parsed into a series of cores and a dynamic programming algorithm is used to align a query sequence to a template.
• NoCore: Both the query sequence and template are treadted as a string of residues. A dynamic programming algorithm is used to align the query sequence to the template.
• IP: Each template is parsed into a series of cores. Pairwise contact potentail is incorporated in the energy function. A linear programming method is used to align the query sequence to the template.

Scoring

The scoring function contains the following items:

• Sequence similarity
• Secondary structure compatibility
• Solvent accessibility
• Pairwise contact potential

Statistical measures for weight training and fold recognition quality assessment:

• After the query sequence is threaded to all the templates, a SVM (Support Vector Machines) technique is used to choose the best template for the query sequence. The SVM output is the predicted alignment accuracy (i.e., the number of aligned positions) of a give sequence-template alignment.
• RAPTOR employs a BLAST-like E-value to evaluate the significance of a sequence-template alignment. The E-value confidence score is easy to be understood.