Comments on "linguistic features in eukaryotic genomes"

Tsonis and Tsonis [1] study rank-ordered distributions of the number of occurrences of protein domains in four different organisms, and they argue that the power-law decay, f ϰ 1/r, of the number f of occurrences of a protein domain with its rank r suggests the presence of linguistic features in eukaryotic genomes, and that this finding " may lead to important clues about the evolution of languages , DNA, and information processing. " We believe that these conclusions are too far-fetched, and that the paper attaches Zipf's law too closely to language alone, as if it is a feature exclusive to human languages. In addition , we would like to mention that (i) the power-law decay of the number of occurrences of protein domains with their rank has been found for dozens of different organisms in Refs. 2 and 3 and that (ii) the ubiquitous presence of those power laws could be explained by simple models of sequence evolution. Admittedly, Zipf's law was first discovered in English, German, and other human languages, but George Kings-ley Zipf himself explored the similar plot in city population, settlement size, and human migration [4]. Today, Zipf's law is known to appear in many contexts unrelated to languages: individual incomes, company sizes, inter-net traffic, web page popularities, number of citations, microarray and gene expression data [5–7], etc. (For an attempt to summarize various observations of Zipf's law, see e.g., [8]). Even randomly generated texts, treating character strings between two blank spaces as " words, " exhibit a rank-frequency statistics consistent with Zipf's law [9]. Can we say that these examples of Zipf's law exhibit " linguistic features? " Clearly not. The size of an earthquake has nothing to do with linguistics, and the text typed by a monkey is not the work of Shakespeare. Yet, both texts and the size distribution of earth quakes are consistent with Zipf's law. Using the language metaphor has a long history in molecular biology, and one of its earliest fruits may be the discovery of the genetic code [10]. However, as powerful as it might be as a metaphor, Zipf's law observed in genomic data is more likely caused by simple processes of sequence evolution mentioned in [2, 3] than by forces that attempt to endow biomolecules with linguistic features. Both Refs. 2 and 3 show that randomly occurring mutations and duplications are sufficient to generate a power-law distribution …