Evaluation of compositional nonrandomness in proteins

Abstract

Cornish-Bowden and Marson have recently suggested that the finite sampling component of Q, a measure of nonrandomness in the amino acid composition of proteins, may have been underestimated because it was calculated on the basis of the genetic code table frequencies rather than on the basis of the average natural abundance with which the twenty amino acids actually occur in proteins. This underestimate would lead to an overestimate of Qc a measure of selective effects above and beyond those imposed by the average natural abundance of the amino acids. In this paper the finite sampling component of Q is quantitatively estimated on the basis of these natural abundances and found to reduce Qc from its previous average value of 24.3 to the lower value of 9.7, with the standard deviation of the population of Qc values being 12.5. Individual Qc values are given for 81 protein families of mean composition per 61 codons of Ala5.3Arg2.4Asn3.0Asp3.6Cys1.5Gln2.6Glu3.5Gly4.7His1.3Ile3.4Leu4.5Lys4.2Met1.0Phe2.3Pro2.3Ser4.2Thr3.6Trp0.8Tyr2.6Val4.2. The mean Qc value of 9.7 is notably small, and indicates that quantitatively minimal adjustments away from the average protein composition are necessary to maintain many different biological functions. This small value, however, is shown to differ significantly from the value of zero expected were the natural abundances of the amino acids the only selective constraint. These small deviations from the natural abundances are thus effectively selected for in the Darwinian sense.