Using Reduced Amino Acid Alphabet and Biological Properties to Analyze and Predict Animal Neurotoxin Protein

Abstract

Background: Because of the high affinity of these animal neurotoxin proteins for some special target site, they were usually used as pharmacological tools and therapeutic agents in medicine to gain deep insights into the function of the nervous system. Background and Objective: The animal neurotoxin proteins are one of the most common functional groups among the animal toxin proteins. Thus, it was very important to characterize and predict the animal neurotoxin proteins. Methods: In this study, the differences between the animal neurotoxin proteins and non-toxin proteins were analyzed. Results: Significant differences were found between them. In addition, the support vector machine was proposed to predict the animal neurotoxin proteins. The predictive results of our classifier achieved the overall accuracy of 96.46%. Furthermore, the random forest and k-nearest neighbors were applied to predict the animal neurotoxin proteins. Conclusion: The compared results indicated that the predictive performances of our classifier were better than other two algorithms.