Implementation

The RCDI/eRCDI program has been implemented in PHP as a web-server and it is freely available as one of the tools of the CAIcal server [10]. The program is also available for local use as a Perl script to calculate the RCDI and eRCDI for a great amount of sequences.

Data input/output

The inputs to calculate the RCDI are: 1) a gene (or set of genes) in FASTA format; 2) the codon usage table of the reference genome in codon usage database [11] format and 3) the genetic code. Other input parameters are related to the calculation of the eRCDI and may be modified according to user requirements. The web-server provides 4 outputs: 1) Genes' parameters: the RCDI values, the frequencies for all codons [(CiFa/CiFh)Ni; see description below] and the %G+C of each gene; 2) Global parameters: the mean %G+C and amino acid composition, and the number of sequences used to calculate the eRCDI; 3) Statistical tests: the RCDI/eRCDI server performs a Kolmogorov-Smirnov test for normality of the eRCDI and a Chi-Square Goodness-of-Fit test to evaluate if the test sequences are homogeneous in %G+C and amino acids composition and 4) Expected RCDI: the mean RCDI value from random sequences and the eRCDI value.

Algorithm

Relative Codon Deoptimization Index (RCDI)

The RCDI/eRCDI server calculates the RCDI by the equation (1), where CiFa is the relative frequency of codon i for a specific amino acid in the test sequence; CiFh is the relative frequency of codon i for a specific amino acid in the reference sequence; Ni is the number of occurrences of codon i in the test sequence; and N, the total number of codons in the test sequence. The RCDI ranges from 1 (the codon usage of the test sequence is completely optimized to the codon usage of the reference genome) to N (increases with the deoptimization of the test sequence). If the reference genome uses all 61 codons, the maximum RCDI value would be equal to the inverse of the lowest synonymous codon frequency value. E.g. for the human genome, the maximum RCDI is 17.64, in the case of a sequence made up with solely TCG codons, which have the lowest synonymous frequency (0.0567).

(1)

Interpretation of the results

RCDI values, based on the Mueller et al. (2006) article [1], may allow for an estimation of the translation rate of viral genes in particular, and the whole genome in general. The higher the similarity between viral and host genes (close to an RCDI value of 1), the higher the translation rate is [1]. The first and most straightforward application is to improve expression of proteins in heterologous expression systems, whether increasing the translation rate or manipulating it to obtain proper folding by choosing a sequence with RCDI close to that of the viral natural host. Furthermore, from a genetic point of view, estimation of RCDI may provide insight into the degree of coevolution between hosts and viruses. A low RCDI might indicate high adaptation to a host. Moreover, a high RCDI might also indicate that some genes are expressed in latency phases or even that the virus might present a low replication rate.

RCDI/eRCDI complements information from other indices

The information gained from the RCDI may be complementary to that of other common indices like CAI [12] or the Effective Number of Codons (ENC) [13]. Figure 1 shows results of RCDI, CAI, ENC and G+C percentage at the third position of the codon (%G+C₃) of 2500 viral sequences (chosen randomly from the NCBI genome resources http://www.ncbi.nlm.nih.gov/sites/genome). CAI and RCDI were calculated using the mean codon usage of the human genome as a reference set. As one may expect, there is a certain correlation (R² = 0.4023) between the ENC and the RCDI (Figure 1a). However, the ENC is a measure of how biased a gene is from an equal use of synonymous codons and it is independent of a reference genome, whereas the RCDI is genome dependent. The indices CAI and RCDI, that are both genome dependent, have a low correlation (R² = 0.0197), reinforcing the idea that these indices may be complementary (Figure 1b). CAI is a measure of codon usage adaptation to the most used synonymous codons of a reference genome [14] and is commonly used to predict highly expressed genes [15]. On the other hand, the RCDI is used to assess if the codon usage of a gene is similar to the codon usage of a reference genome. In addition figure 1c shows that RCDI values and %G+C₃ are linearly independent.

Both indices, CAI and ENC can be associated with an expected value, the expected CAI by generation of random sequences [9] and an estimated ENC by a formula that uses the %G+C at the third position of the codon [13]. In this article we introduce the eRCDI as a measure to assess the significance of the RCDI value. Thus, based on the RCDI and eRCDI, it is possible to estimate if a gene is deoptimized and if this deoptimization is the product of compositional biases (%G+C and/or amino acid composition).