BACKGROUND In the process of protein construction, buried hydrophobic residues tend to assemble in a core of a protein. Methods used to predict these cores involve use or no use of sequential alignment. In the case of a close homology, prediction was more accurate if sequential alignment was used. If the homology was weak, predictions would be unreliable. A hydrophobicity plot involving the hydropathy index is useful for purposes of prediction, and smoothing is essential. However, the proposed methods are insufficient. We attempted to predict hydrophobic cores with a low frequency extracted from the hydrophobicity plot, using wavelet analysis. RESULTS The cores were predicted at a rate of 68.7%, by cross-validation. Using wavelet analysis, the cores of non-homologous proteins can be predicted with close to 70% accuracy, without sequential alignment. BACKGROUND The program used in this study is available from Intergalactic Reality (http://www.intergalact.com). BACKGROUND hirakawa@grt.kyushu-u.ac.jp, kuhara@grt.kyushu-u.ac.jp