The prediction of protein structure is important as the function of a protein is determined by its structure. The availability of large families of homologous sequences revolutionized secondary structure prediction. This poster presents an Artificial Neural Network-based computational framework for predicting secondary structure of the protein using biologically meaningful position specific scoring matrix (PSSM) as a feature. The PSSM matrix is obtained from homologous sequences using the PSI-BLAST. Evaluating protein structure predictions, however, has many elements of subjectivity, making it difficult to quantify the results that are significant. Empirical analyses are being conducted to evaluate the performance of the proposed approach using metric measures such as precision, recall, F measure.