Adaptation studied with the self-consistent codon index: genomic spaces, metabolic network comparison, minimal gene sets and viral classification

Facts and ideas presented in this short review concern some recent developments at the interface between microbial spaces, metabolic network comparison, minimal gene sets and viral classification. The guiding line to all results presented here is to derive biological information from genome sequences by means of a purely statistical analysis and an appropriate design of algorithms. The paper is an updated version of (Carbone 2005). 1 Some background and motivation Proteins are formed out of 20 amino-acids which are coded in triplets of nucleotides, called codons. The four nucleotides (A, T, C, G) define 64 codons used in the cell. Codons are not uniformly employed in the cell, but at the contrary, certain codons are preferred and we speak about codon bias. There are several kinds of codon biases and some of them are linked to specific biological functions. Statistical analysis of DNA sequences and in particular of codon bias were performed from the moment that long chunks of DNA sequences were publicly available in the early eighties (Grantham et al. 1980; Wada et al. 1990), and the roots for these studies can be traced back to the sixties (Sueoka 1962; Zuckerkandl and Pauling 1965). However with the increasing number of bacterial genome sequences from a broad diversity of species, this field of research has been revivified in the last few years (Koonin and Galperin 1997; Lin and Gerstein 2000; Radomski and Slonimski 2001; Knight et al. 2001; Sicheritz-Ponten and Andersson 2001; Daubin et al. 2002; Lin et al. 2002; Lobry and Chessel 2003; Sandberg et al. 2003; Jansen et al. 2003). Biased codon usage may result from a diversity of factors: GC-content, preference for codons with G or C at the third nucleotide position (Lafay et al. 1999), a leading strand richer in G+T than a lagging strand (Lafay et al. 1999), horizontal gene transfer which induces chromosome segments of unusual base composition (Moszer et al. 1999), and in particular, translational bias which has been frequently noticed in fast growing prokaryotes and eukaryotes (Sharp and Li 1987; Sharp et al. 1986; Medigue et al. 1991; Shields and Sharp 1987; Sharp et al. 1988; Stenico et al. 1994). Three main facts support the idea of ”translational impact”: highly expressed genes tend to use only a limited number of codons and display a high codon bias (Grantham et al. 1980; Sharp and Li 1987), preferred codons and isoacceptor tRNA content exhibit a strong positive correlation (Ikemura 1985; Bennetzen and Hall 1982; Bulmer 1987;