Currently I am working on automatic feature extraction for protein-protein interactions. My study includes why protein interacts and how to predict the interaction within protein to understand the biological fuction performed by different proteins. Different type of preprocessing and freature selction method is also a part of my current study. I work on different type of protein properties like disolvation energy, solvent accessible surface area, conservation score, electrostatic energy, hydrophobicity etc. Differenty type of proteins like homo-hetero, obligate-transient, polar-nopolar, hydrophobic-nonhydrophobic. I am working to find best feature sets to differentiate those type of proteins pairs. The accuracy of any classification lies on the input feature vector. So, I am working to find an automated process which will generate the best feature sets based on different property of protein so that the classifier can differentiate different types of proteins.
A part of my study is to understand the four types of structure of protein ( primary, secondary, tertiary and quaternary ). The idea of finding best features for classificaton will help to identify the types of different proteins and their compositions and the functions that they can perform.