Computer-based prediction of mitochondrial-targeting peptides

Computational methods are invaluable when protein sequences, directly derived from genomic data, need functional and structural annotation. This is the case of subcellular localisation, a feature necessary for understanding the protein role and the compartment where the mature protein is active. Mitochondrial proteins encoded on the cytosolic ribosomes carry specific patterns in the precursor sequence from where it is computationally possible to recognise a peptide targeting the protein to its final destination. Here we discuss to which extent it is feasible to develop computational methods for detecting targeting peptides in the precursor sequences and benchmark our and other methods on the human mitochondrial proteins endowed with an experimentally characterised targeting peptide. Furthermore we illustrate our web server and its usage on the human poorly characterised proteome in order to infer targeting peptides, their cleavage sites and whether the targeting peptide regions contains or not arginine rich recurrent motifs. By this, we add some other 2800 human proteins to the 124 ones already experimentally annotated with a targeting peptide.