Prediction of Disulfide Connectivity Patterns from Protein Sequence

A new computational method is introduced to predict disulfide connectivity patterns in a protein chain, starting with the assumption that the disulfide bonding state of each cysteine is known. The method uses support vector machines based on the pairwise local similarities between cyteineneighboring sequences and the distance between the cysteine pair under consideration. According to the experimental results over a common data set, the new system provides improved prediction accuracy in comparison with existing methods.