A method of semiempirical identification of structural domains is proposed. The procedure is based on the comparison of amino acid sequences in groups of homologous proteins. This approach was tested using 32 known protein sequences from different cytochrome b5, cytochrome c, lysozyme, hemoglobin, and myoglobin proteins. The method presented was able to identify all structural domains of these reference proteins. A consensus secondary structure provided information on structural content of these domains predicting correctly 21 of 23 (91%) of alpha-helices. We applied this method to six homologous phytochrome sequences from Avena, Arabadopsis, Cucurbita, Maize, Oryza, and Pisum. Some of the identified domains can be assigned to the known tertiary structure categories. For example, an alpha/beta domain is localized in the region known to stabilize the phytochrome chromophore in the red light absorbing form (Pr). One alpha-helical and one alpha/beta domains are localized in regions important for the chromophore stabilization in the far-red absorbing form (Pfr). From an analysis of noncovalent interaction patterns in another domain it is proposed that a phytochrome dimer contact involves two segments localized between residues 730 and 821 (using numbering of aligned sequences). Also, a possible antiparallel beta-sheet structure of this region has been suggested. According to this model, the long axis of the interacting structures is perpendicular to a twofold symmetry axis of the phytochrome dimer.