Pripper: prediction of caspase cleavage sites from whole proteomes

Abstract

Background: Caspases are a family of proteases that have central functions in programmed cell death (apoptosis) and inflammation. Caspases mediate their effects through aspartate-specific cleavage of their target proteins, and at present almost 400 caspase substrates are known. There are several methods developed to predict caspase cleavage sites from individual proteins, but currently none of them can be used to predict caspase cleavage sites from multiple proteins or entire proteomes, or to use several classifiers in combination. The possibility to create a database from predicted caspase cleavage products for the whole genome could significantly aid in identifying novel caspase targets from tandem mass spectrometry based proteomic experiments.

Results: Three different pattern recognition classifiers were developed for predicting caspase cleavage sites from protein sequences. Evaluation of the classifiers with quality measures indicated that all of the three classifiers performed well in predicting caspase cleavage sites, and when combining different classifiers the accuracy increased further. A new tool, Pripper, was developed to utilize the classifiers and predict the caspase cut sites from an arbitrary number of input sequences. A database was constructed with the developed tool, and it was used to identify caspase target proteins from tandem mass spectrometry data from two different proteomic experiments. Both known caspase cleavage products as well as novel cleavage products were identified using the database demonstrating the usefulness of the tool. Pripper is not restricted to predicting only caspase cut sites, but it gives the possibility to scan protein sequences for any given motif(s) and predict cut sites once a suitable cut site prediction model for any other protease has been developed. Pripper is freely available and can be downloaded from http://users.utu.fi/mijopi/Pripper.

Conclusions: We have developed Pripper, a tool for reading an arbitrary number of proteins in FASTA format, predicting their caspase cleavage sites and outputting the cleaved sequences to a new FASTA format sequence file. We show that Pripper is a valuable tool in identifying novel caspase target proteins from modern proteomics experiments.

Figure 1

The ROC curves of the trained classifiers. Area under ROC curve (AUC) is highest for SVM-6-4 and lowest for J48-4-2, and thus SVM-6-4 and RF-12-12 show better performance than the J48-4-2 classifier.

Figure 2

ROC curve comparisons of caspase cleavage site classifiers. The data points represent the average True positive and False positive rates from the 10 000 bootstrap values. The lines surrounding the data points represent the standard deviation calculated from the bootstrap data. The ROC curve shows that the Vote classifier that combined the RF-12-12 and SVM-6-4 classifiers appears to be the best among the compared classifiers, since its True positive rate is very close to one and False positive rate is zero.

Figure 3

Caspase cleavage products identified based on MS/MS-data. Protein identifications were done with Mascot against a database of predicted caspase cleavage products created with Pripper (Vote-classifier: SVM-6-4 and RF-12-12). Complete protein sequences are shown with all the identified peptides (red), caspase cleavage motifs (yellow) and peptides identified at the cleavage site (black box). Peptide spectrum matches for the cleavage site peptides are also shown. A) Cytokeratin 18 identified from the 2-DE sample B) Potential new caspase target leukosialin identified from the iTRAQ-sample.