The paper reports results obtained from a study on database validation concerning a leucine zipper motif, utilizing both a functional motif model and a structural motif model. As an example for this method, a leucine zipper motif is chosen, which is a subsequence consisting of two twisted alpha helix sequences and preceded by a DNA binding site. For a functional motif model, an HMM (hidden Markov model) which is trained with leucine zipper subsequences is employed. For a structural motif model, a neural network which is trained to classify an alpha helix region, is employed. Because only 122 such leucine zipper sequences are an the Swiss Protein database (R.22), there is a possibility that the HMM could not learn the general mechanism for helical structures completely. Therefore, a structural motif model for an alpha helix is utilized to eliminate non-helical sequences. Fortunately, there are numerous secondary structures' examples available in the PDB database. All polypeptides in the Swiss Protein database v.22 are examined with a combination of an HMM and a neural network. For predicting a leucine zipper region, an HMM achieved 92.301 percent and improved up to 92.411 percent by combining a neural network.