tAI is an R package (and two perl scripts) for the analysis of codon usage in DNA coding sequences. It implements the tRNA adaptation index (tAI) described in dos Reis et al. (2003, 2004). The tAI package is distributed WITHOUT WARRANTY under the terms of the GNU General Public License. See the file LICENSE for details.
This package is basically the same I used in my papers on tAI, with just a few minor modifications to make it a little friendlier. If you have any doubts, questions, bug reports, etc. please contact me at:
Mario dos Reis
School of Biological and Chemical Sciences
Queen Mary University of London
From the R prompt (assumming you have the devtools package installed) type
> devtools::install_github("mariodosreis/tai")
And that's it! The package has been installed!
You are assumed to know how to use a shell or command prompt, as well as
some basic knowledge of R. $ denotes the shell prompt (so don't type
it in!).
The tRNA adaptation index (tAI) measures the degree of co-adaptation
between a coding sequence and the tRNA pool of an organism (dos Reis 2003, 2004). File
misc/ecolik12.ffn contains 49 coding sequences extracted from the genome
of Escherichia coli K-12. We can use the files in the package to
calculate tAI for each one of the genes contained in this file. First we
need to calculate the frequencies of the 61 coding codons for every
sequence. Go to the misc directory in this package and
type (note that perl must be installed in your system and in your path):
$ perl codonM ecolik12.ffn ecolik12.m
The file ecolik12.m contains the output of the codonM script: a
matrix of codon frequencies per ORF. It should look like:
11 19 10 13 11 10 6 9 ...
6 4 3 6 0 6 1 3 ...
13 11 5 12 0 2 3 5 ...
0 0 2 0 1 1 1 0 ...
8 8 2 6 0 4 2 1 ...
...
Each row represents one ORF or gene (in our case, there should be 49
rows in the file ecolik12.m), and the columns represent each one of
the 61 coding codons, arranged in this fashion:
1 TTT
2 TTC
3 TTA
4 TTG
5 TCT
6 TCC
7 TCA
8 TCG
9 TAT
10 TAC
11 TGT
12 TGC
13 TGG
14 CTT
15 CTC
16 CTA
17 CTG
18 CCT
19 CCC
20 CCA
21 CCG
22 CAT
23 CAC
24 CAA
25 CAG
26 CGT
27 CGC
28 CGA
29 CGG
30 ATT
31 ATC
32 ATA
33 ATG
34 ACT
35 ACC
36 ACA
37 ACG
38 AAT
39 AAC
40 AAA
41 AAG
42 AGT
43 AGC
44 AGA
45 AGG
46 GTT
47 GTC
48 GTA
49 GTG
50 GCT
51 GCC
52 GCA
53 GCG
54 GAT
55 GAC
56 GAA
57 GAG
58 GGT
59 GGC
60 GGA
61 GGG
Notice that STOP codons have been excluded. codonM, also ignores the first codon in every sequence, this is because it is always a Methionine codon (even if its not coded by the canonical ATG). The codons above follow the TCAG ordering. The standard genetic code ordered this way is
AAs = FFLLSSSSYY**CC*WLLLLPPPPHHQQRRRRIIIMTTTTNNKKSSRRVVVVAAAADDEEGGGG
Starts = ---M------**--*----M---------------M----------------------------
Base1 = TTTTTTTTTTTTTTTTCCCCCCCCCCCCCCCCAAAAAAAAAAAAAAAAGGGGGGGGGGGGGGGG
Base2 = TTTTCCCCAAAAGGGGTTTTCCCCAAAAGGGGTTTTCCCCAAAAGGGGTTTTCCCCAAAAGGGG
Base3 = TCAGTCAGTCAGTCAGTCAGTCAGTCAGTCAGTCAGTCAGTCAGTCAGTCAGTCAGTCAGTCAG
Now start R in the same directory you have been working. (> indicates
the R prompt). Type:
> require("tAI")
This will load the necessary functions into R. The file
ecolik12.trna contains the gene copy number of every kind of tRNA in
the E. coli K-12 genome. We need this information in order to
calculate tAI:
> eco.trna <- scan("ecolik12.trna")
This file contains 64 rows, corresponding to the anticodon complements of each tRNA species (e.g. if the tRNA anticodon is GAA, the complement is TTC). The tRNAs are ordered according to their anticodon complement, in the same order as in codonM's output as indicated above, but with added STOP codons. STOP codons are at positions 11, 12 and 15 as in the standard genetic code diagram above. Now we can calculate the relative adaptiveness values for each codon in the E. coli genes:
> eco.ws <- get.ws(tRNA=eco.trna, sking=1)
Now, lets load the output of codonM into R:
> eco.m <- matrix(scan("ecolik12.m"), ncol=61, byrow=TRUE)
We will ignore Methionine codons in our analysis (there is no automatic way to differentiate between 'START' Met-tRNA genes and normal Met-tRNAs in any genome):
> eco.m <- eco.m[,-33]
Now we can finally calculate tAI:
> eco.tai <- get.tai(eco.m, eco.ws)
> hist(eco.tai)
The last command will plot an histogram of the tAI values. Highly expressed genes present high tAI values (> 0.4), which means that their codon usage resembles the genomic structure of tRNA genes.
Now a good question arises, how much of the total codon usage of these
genes (in terms of Nc) is due to adaptation to the tRNA gene pool? In
order to answer this question, we need to calculate Nc for every
sequence in ecolik12.ffn. You should have codonW (or other codon usage
package able to calculate Nc) installed. If you do have codonW, and it
is in your path, you can use codonZ to quickly and efficiently compute
a set of codon usage statistics. At the shell type:
$ codonZ ecolik12.ffn ecolik12.w
If you don't have codonW installed, file ecolik12.w is already provided
in the misc directory. Now from the R prompt:
> df <- read.table("ecolik12.w", header=TRUE)
If you are using the codonW version distributed by me, the above command should work fine, if you are working with the original distribution, you should type instead:
> df <- read.table("ecolik12.w", header=TRUE, na.strings = "*****")
Lets plot the relationship between tAI and Nc:
> plot(eco.tai ~ df$Nc)
As you can see, genes with very low Nc values (highly biased), correlate with high tAI values (highly co-adapted to the tRNA gene pool).
> cor(eco.tai, df$Nc, use="p")
You should get a value of –0.9100338
Formally, we can calculate the correlation between tAI, and the corrected Nc, f(GC3s) – Nc, we call this correlation S, because it reflects the intensity of translational selection acting on our sample of genes:
> eco.s <- get.s(eco.tai, df$Nc, df$GC3s)
You should get a value of 0.9065442
For more details, read the references!
[1] dos Reis M, Savva R, and Wernisch L. (2003) Unexpected correlations between gene expression and codon usage bias from microarray data for the whole Escherichia coli K-12 genome. Nucleic Acids Research, 31: 6976–6985.
[2] dos Reis M, Wernisch L, and Savva R. (2004) Solving the riddle of codon usage preferences: a test for translational selection. Nucleic Acids Research, 32: 5036–5044.