Intensive sequence comparisons to predict protein secondary structures. Integration into a software package: ANTHEPROT

The combination of structural class assignment with intensive sequences comparisons has permitted to develop a new concept in the prediction of the secondary structure of proteins called a self optimised prediction method (SOPM). The accuracy of this method has been checked onto an updated release of the Kabsch & Sander database which comprises 239 protein chains. This new method correctly predicts more than 69% of amino acids for a 3-states description of the secondary structure (a helix, j3 sheet and coil) in the whole database (46,223 amino acids). The correlation CoeJticients are Ca=O53, C/3=0.51 and Ccoil=O.49. By also considering the B turn state the success is 61% with Ca=O.53, C#I=OSl, Ct= 0.27 and Ccoit=O.40. RMSD as tow as 10% in the secondary structure content are obtained. This method has been integrated into a software package catted ANTHEPROT in order to facilitate the comparison with others methodr. Programs are available by ftp to ibcp fr.