PRECIS

See the actual review in Comp. Rev. Here, we overview it.
The review (See CR as cited above) begins with problem and its importance to assessing structure and ultimately function of proteins. The problem has been researched before and this paper points to previous results and the authors' goals of enriching models. A Support Vector Machine approach postulates a first stage to operate as a classifier to separate proteins into three categories based on whether cysteines appear in all, none, or some of the disulfide bridges. The second stage employs a binary classifier to determine individual cysteine states. The model mathematics and matters such as simplifying assumptions are laid out carefully. "Experiments" with SVM models with differing kernel types (linear, polynomial, and radial bias functions) are outlined and results (accuracy of 85%) are presented for five classifiers' performance on a set of over seven hundred proteins (selection details presented). The paper is heavily referenced on the biology and on the SVM fronts, presenting a probable challenge for readers who have not been following the topics rather closely.