MITOPRED

MITOPRED predicts nucleus-encoded mitochondrial proteins from protein sequences using Pfam domain occurrence patterns and amino acid compositional differences to distinguish mitochondrial and non-mitochondrial proteins.

Key Features:

• Algorithmic basis: Uses Pfam domain occurrence patterns and amino acid composition differences between mitochondrial and non-mitochondrial proteins to identify mitochondrial-targeted proteins, including those lacking conventional targeting signals.
• Prediction accuracy: Reports 100% specificity at 44% sensitivity and 67% specificity at 99% sensitivity, with a Matthews correlation coefficient of 0.73 across eukaryotes, outperforming TargetP (0.51) and PSORT (0.53).
• Pre-calculated proteome predictions: Provides pre-computed predictions for Saccharomyces cerevisiae, Caenorhabditis elegans, Drosophila, Homo sapiens, Mus musculus, Arabidopsis, and for eukaryotic sequences in Swiss-Prot and TrEMBL.

Scientific Applications:

• Genome-wide mitochondrial protein annotation: Enables estimation of nucleus-encoded mitochondrial protein complements in complete genomes, for example predicting 1,362 mitochondrial proteins in Homo sapiens (~4.8% of the proteome).
• Mitochondrial function and disease studies: Supports studies of mitochondrial function and the role of mitochondrial proteins in biological processes and diseases across diverse eukaryotic species.

Methodology:

Classification is based on an algorithm using Pfam domain occurrence patterns and amino acid compositional differences between mitochondrial and non-mitochondrial proteins.