Comparative Analysis of Disulfide Bond Determination Using Computational-Predictive Methods and Mass Spectrometry-Based Algorithmic Approach

Identifying the disulfide bonding pattern in a protein is critical to understanding its structure and function. At the state-of-the-art, a large number of computational strategies have been proposed that predict the disulfide bonding pattern using sequence-level information. Recent past has also seen a spurt in the use of Mass spectrometric (MS) methods in proteomics. Mass spectrometry-based analysis can also be used to determine disulfide bonds. Furthermore, MS methods can work with lower sample purity when compared with x-ray crystallography or NMR. However, without the assistance of computational techniques, MS-based identification of disulfide bonds is time-consuming and complicated. In this paper we present an algorithmic solution to this problem and examine how the proposed method successfully deals with some of the key challenges in mass spectrometry. Using data from the analysis of nine eukaryotic Glycosyltransferases with varying numbers of cysteines and disulfide bonds we perform a detailed comparative analysis between the MS-based approach and a number of computational-predictive methods. These experiments highlight the tradeoffs between these classes of techniques and provide critical insights for further advances in this important problem domain.