New Mitochondrial Reference Set Available - IMPI

MitoMiner now includes a new reference set - IMPI (Integrated Mitochondrial Protein Index) for Human, Mouse and Rat.

The IMPI reference set is a collection of proteins with strong evidence for mitochondrial localisation developed using machine learning. Many clues to mitochondrial localisation exist such as targeting sequences as well as direct evidence for mitochondrial localisation such as mass spectrometry and antibody staining. As no one indicator is as as accurate as all indicators combined, IMPI uses two machine learning techniques to objectively combine all evidence available for a protein. This includes the mass spectrometry and GFP tagging experimental data from the studies included in MitoMiner, targeting sequence predictions from 3 programs, and antibody staining from the Human Protein Atlas. The machine learning then selects proteins that have similar properties and evidence to characterised mitochondrial proteins whilst discarding those most similar to known non-mitochondrial proteins. The final list combines the list of characterised mitochondrial proteins with those from the machine learning to produce the IMPI reference set. This provides a comprehensive dataset and solves the problem of deciding an arbitrary threshold for what level of evidence can be considered "mitochondrial".