Mitochondria consist of several hundreds of proteins, the vast majority of which are synthesized in the cytosol as precursor proteins from where they are targeted to and imported into mitochondria. The transport of proteins into mitochondria relies on specific targeting information encoded within the protein sequence, known as mitochondrial targeting sequences (MTSs). These N-terminal extensions are usually between 8 and 80 residues long and form amphipathic helices with one hydrophobic and one positively charged surface. Receptors on the mitochondrial surface recognize the MTSs and direct precursors through protein-conducting channels in the outer and inner membrane to the mitochondrial matrix, where presequences are often removed by proteases. In addition to these MTSs, many mitochondrial proteins contain internal matrix targeting sequences (iMTSs) which share the same structural features with MTSs. These iMTSs are neither necessary nor sufficient for mitochondrial targeting, however, they help to increase the import-competence of precursor proteins as they bind to the TOM receptors and presumably facilitate the unfolding of precursors on the mitochondrial surface. Prediction algorithms allow the identification of iMTSs in protein sequences. In this chapter, we present iMLP, an agnostic algorithm for the prediction of iMTS propensity profiles. This iMTS prediction tool is provided via an iMLP webservice at http://iMLP.bio.uni-kl.de and is also available as a BioFSharp application that can be executed locally. We describe and explain the usage of this prediction algorithm and how to interpret the results of this valuable tool.