| Knowing the subcellular location of a protein is an important step in understanding its biological functions.The Golgi apparatus is an important organelle and can be found in most eukaryotic cells.The defects of Golgi apparatus are associated with some kinds of neurodegenerative diseases such as Alzheimer disease and Parkinson’s disease.Knowing the type of a Golgi-resident protein is an important step in understanding its association with these diseases as well as its role in biological processes.In this paper,we developed a new method to identify whether a protein is a Golgi-resident protein or not in plant cells.We proposed to incorporate transmembrane domain information and six different kinds of physicochemical properties of amino acids in the general form of Chou’s pseudo-amino acid compositions.By using SVM based classifiers,our method achieved over 90% prediction accuracy in a 5-fold cross validation,which is much better than the other state-of-the-art methods.It is not enough for us to just identify the Golgi-resident proteins,we should also understand the type of a Golgi-resident protein.Knowing the type of a Golgi-resident protein is an important step in understanding its molecular functions.The Golgi apparatus consists of two main parts: cis-Golgi network(CGN),trans-Golgi network(TGN).The cis-Golgi network,facing the endoplasmic reticulum,receives the biosynthetic output which comes from the endoplasmic reticulum.Different kinds of Golgi-resident proteins play different roles in biological procedures.In this paper,we proposed a SVM-based prediction model which combined Chou’s pseudo amino acid compositions with the positional specific physicochemical properties and made use of the feature selection framework based on mutual information simultaneously.In leave-one-out cross-validation test,91.24% prediction accuracy was achieved by our method with only 49 features,which is better than all the state-of-the-art ones.This result demonstrates that our computational model can be useful in discriminating the types of Golgi-resident proteins.To make our methods more reasonable,we also carried out some other kinds of comparisons.As demonstrated in the prediction results,our methods perform better than all the state-ofthe-art methods.Besides,we also made a comparison in the field of feature selection methods.The results show that our method has the potential to be applied in predicting a wide range of protein attributes and perform stably in predicting sub-Golgi locations of proteins.In our paper,we proposed a systematic method to analyze Golgi-resident proteins which includes two main steps.First,to identify whether a protein is a Golgiresident protein in plants or not.Second,to find out the type of a Golgi-resident protein.As far as we know,our methods perform better than all the other methods in the same field. |
