Auto-Coding of Amino Acid Sequences for Prediction of Protein Tertiary Structure

Abstract

Protein structure prediction is turning out to be one of the major challenges in the field of
bio-informatics. It is highly important in medicine, especially in drug design and
biotechnology. Proteins, being the basic building unit of all organisms, require experimental
techniques for prediction of related structures. Among available methods, soft-computational
tools provide readily available solutions for making predictions with less complexity, higher
reliability and less time. The Artificial Neural Network (ANN) is one such tool which is used
for structure prediction of proteins. This method is a machine learning approach in which
ANNs are trained to make them capable of recognizing the 8-level subclasses of secondary
structure. After the subclasses are recognized in a given sequence, their association with 3-
level secondary protein structures is derived. The final structure is obtained from a majority
selection from the protein structure. The work is also done in the reverse way, by predicting
the 3-level secondary structure from the primary structure. This is done to confirm the
accuracy of the prediction. In this work, ANNs are used as classifier to predict the secondary
structure.