TPpred 2.0

TPpred 2.0 predicts mitochondrial targeting peptides and their cleavage sites in protein sequences to identify N-terminal sorting signals for mitochondrial and plastidic protein import.

Key Features:

• Machine learning approach: Employs Grammatical-Restrained Hidden Conditional Random Fields (GR-HCRFs) to analyze protein sequences for targeting peptides and cleavage sites.
• Improved cleavage-site prediction: Reports average cleavage-site errors of 7 residues for mitochondrial proteins and 15 residues for plastidic proteins.
• Performance metrics: Achieves reported accuracy of 96%, a Matthews correlation coefficient of 0.58, and a false-positive rate of 3.0%.
• Correlation with experimental data: Predictions correlate with experimental subcellular localization annotations when available.

Scientific Applications:

• Proteome annotation: Identification of proteins containing mitochondrial or plastidic targeting peptides for genome- and proteome-scale annotation.
• Inference of mature protein sequences: Prediction of cleavage sites to infer sequences of mature mitochondrial and plastidic proteins.
• Comparative proteomics: Estimation of the proportion of sequences with targeting peptides in human, Arabidopsis thaliana, and yeast proteomes to study import mechanisms.

Methodology:

Uses Grammatical-Restrained Hidden Conditional Random Fields (GR-HCRFs) machine-learning models to analyze protein sequences and predict presence and cleavage sites of targeting peptides.