Bharanidharan et al. [Biochem. Biophys. Res. Commun. 315 (2004) 1097-1103] claimed that the frequencies of most amino acids are determined by the dinucleotide composition of the genome. Here, regarding a methodological problem in their work, it is suggested that the standard deviations of amino acid frequencies should be determined to indicate how significant a certain deviation from the predicted frequency is. Furthermore, using a different method that is expected to be more reliable, we suggest that the dinucleotide composition cannot explain the observed frequencies of most amino acids, and the deviations of amino acid frequencies from what dinucleotide composition predicts are larger than to be expected by chance. (C) 2004 Elsevier Inc. All rights reserved.