A Semi-Automated Structural Class Dependent Method for the Prediction of Protein Secondary Structures

ABSTRACT

The use of amino acid frequencies for the prediction of protein secondary structures has been limited due to emergence of various mathematical models adding to the quality of predictions. However, these models carry some extra algorithmic burden and are computationally demanding. The objective of the current work is to present an easy to use method based directly on the amino-acid frequencies for prediction of secondary protein structures. The advantage of this method is that a prediction is presented in the form of six possible readings of the amino acid sequence. This provides the opportunity to the user to consider the general secondary structure content of the protein, instead of returning a particular outcome as most of the widely used prediction implementations do. The results of the method are comparable to the well known software solutions for secondary structure prediction, based on amino acid frequencies, and there is visible potential for improvement of our method. This leads to the conclusion that the frequency approach is still applicable to this particular prediction problem, and can be useful when dealing with novel sequences where homology based approaches have less or nothing to work with.

Keywords:

• secondary protein structure
• prediction method
• amino acid frequencies