The basic pattern of amino acid compositions obtained from the amino acid analysis of whole cell lysates is quite often similar between bacteria and mammalian cells, even though their cells are constructed from various kinds of protein in different quantities. In addition, the amino acid compositions calculated from complete genome are similar among various organisms, assuming conveniently that all genes are equally expressed in cells; this amino acid composition resembles the cellular amino acid composition obtained from amino acid analysis. This coincidence of these two data sets based on completely different principles puzzled us for a long period.  However, this puzzle has been recently solved by understanding that the genome is constructed from putative gene assembly units encoding similar amino acid compositions consisting of 3,000-7,000 amino acid residues; the amino acid compositions based on these small units can be observed in any part of the genome. Thus, the amino acid composition based on a certain gene assembly is independent of its location on the genome, but depends on species. Applying this principle to whole cells consisting of various proteins in different quantities, the cellular amino acid composition shows a species-specific pattern to be precise, but its basic pattern is roughly similar among various organisms. Codon alternations occur synchronously over the genome in biological evolution; as a result, genome structure is quite homogeneous based on codon usages.