Computational prediction of toxic protein

We present a novel computational method to predict a protein as it is toxic or not, from sequence only. The main challenge lies in the fact that neither there is any specific feature nor there is any range of values in the respective structural features supported by all toxic proteins. To resolve this challenge, we have proposed and implemented a new approach. Here, at first we collect a large amount of proteins from different famous and authentic protein database. Then we find out those parameters that toxic proteins have in common in terms of their physical and chemical properties. Only the sequence of amino acid and structural characteristics are used to predict the features. Then we construct a set of criteria from the set of features to be used to predict them such as iso-electric point(PI), aliphatic index, amino acid content, half life, grand average, instability index etc. After constructing features we estimate a range for each feature and establish them through hypothesis testing using statistical distribution. Our study demonstrates that this method can provide an easy and high detection rate of toxic protein.