Prediction of protein subcellular locations using Markov chain models

A novel method was introduced to predict protein subcellular locations from sequences. Using sequence data, this method achieved a prediction accuracy higher than previous methods based on the amino acid composition. For thre e subcellular locations in a prokaryotic organism, the overall prediction a ccuracy reached 89.1%. For eukaryotic proteins, prediction accuracies of 73 .0% and 78.7% mere attained within four and three location categories, resp ectively. These results demonstrate the applicability of this relative simp le method and possible improvement of prediction for the protein subcellula r location. (C) 1999 Federation of European Biochemical Societies.