Predict protein secondary structure

ebisu group

chemogenomix © 2012

Read more below


PSIPRED is a popular structure prediction method that incorporates two feed-forward neural networks to perform an analysis of results obtained by the PSI-Blast homology search algorithm (Altschul et al., 1997)

The method is well validated. At the time of original publication, a stringent cross-validation procedure showed the method to be capable of achieving an average Q3 score of 76.5% and improvements have been made since that benefit from the increase in available data (as have all algorithms that use homology).

At the CASP3 assessment PSIPRED ranked first of all the secondary structure prediction methods evaluated and at the CASP4 assessment was in a set of high performers achieving over 80% on the smaller set of the test examples. In subsequent assessments many groups have used the software as a component of various techniques.

The PSIPRED program continues to be maintained and is available from David Jones group. There is also a web server for users to get predictions for single sequences.

PSIPRED can be run from the PSIPRED prediction server