RSCP: a web server for redox-sensitive cysteine prediction

Introduction Prediction Download Help

What are redox-sensitive cysteines?

Reactive oxygen species (ROS) has been regarded as unwanted by-product of aerobic metabolism. However, under normal conditions, ROS could modify the structure and function of proteins in defined ways. ROS could also act as important signalling molecules in various cellular processes. Cysteine thiol groups of proteins are particularly susceptible to oxidation by ROS/RNS and other electrophilic molecules, and their reversible oxidation is of critical roles for the redox regulation of proteins.

Redox-sensitive cysteines hereby stands for these cysteines which could have changed redox status under different condition. Many of redox-sensitive cysteines were confirmed to be functionaly important.

Why predict redox-sensitive cysteines?

Traditionally, redox-sensitive cysteines are identified by case-by-case studies. With the development of proteomics techniques, it is possible to identify hundreds of redox-sensitive proteins in one single experiment. However, the proteomic techniques are still limited by protein abundance and other technical problems.

In the past years, several computational tools have been developed for predicting redox-sensitive cysteines; however, those methods either only focus on catalytic redox-sensitive cysteines, or depend on structural data. Thus, it's necessary to develop an efficient sequence-based method that can predict both catalytic and non-catalytic redox-sensitive cysteines.Computational prediction of redox-sensitive cysteines could guide further experimental validation and faciliate our understanding of thiol-based redox regulation.

Introduction of RSCP

Based on a newly collected dataset (RSC758) for various types of redox-sensitive cysteines, we identified several types of sequence-based features, including PSSM profile, secondary structure, and the sequential distance to nearby cysteines that can be used for redox-sensitive cysteine prediction. After further feature selection using SVM-RFE, we developed a SVM based tool for redox-sensitive cysteine prediction. Using 10-fold cross-validation, it could achieve overall accuracy of 67.9%, sensitivity of 0.602, specificity of 0.756, MCC of 0.362, and AUC of 0.727. It also achieved robust performance when applied to an independent dataset.

RSCP is the web server implement of this approach. It takes primary sequence as input, and predict the occurrence of redox-sensitive cysteines within the sequence.


Reference:

Sun et al. Prediction of redox-sensitive cysteines using sequential distance and other sequence-based features. Manuscript submitted.



Contact: djguo@cuhk.edu.hk or mingansun@gmail.com