Prediction of mitochondrial proteins of malaria parasite using improved hybrid method and reduced amino acid alphabet

The rate of human death and morbidity due to malaria is increasing in many parts of the developing countries. Thus, there is a great need to understand the critical pathways in malaria parasite in order to develop effective drugs and vaccines. In this work, based on the measure of diversity definition, we introduce the increment of diversity fusion (IDF), an improved hybrid method to predict mitochondrial proteins of malaria parasite. We conduct our experiment on an expanded protein dataset where we require the pairwise identity between two proteins is less than 25%. By choosing amino acids composition as the only input vector, we are able to achieve 65.4% accuracy with 0.32 Mathew's correlation coefficient (MCC) for the jackknife test. Further, incorporting the compositions of the N-terminal and C-terminal regions into the input vector, we show that the prediction results are improved to 82.0% accuracy with 0.64 MCC in the jackknife test. In addition, by combining with the several reduced amino acid alphabet and the hydropathy distribution along protein sequence, we achieve maximum 83.4% accuracy with 0.67 MCC in the jackknife test by using the 64 dipeptide compositions of the reduced amino acid alphabet obtained from Protein Blocks method.