Simulation of Prokaryotic Genome Evolution Subjected to Mutational Pressures Associated with DNA Replication - Biomedical Engineering Systems and Technologies - page 130

positions on the prokaryotic chromosomes because they, as also other essential genes
for cell functioning, tend to be coded in the leading strand [26-28].
The previous computer simulation [21-24] included amino acid composition of the
gene products as the selection constraint whereas here we have presented the results of
other type of simulations in which selections against stop codon occurrence and for the
preservation of the coding signal were applied. The selection for the coding signal was
calculated by the algorithm for recognition of protein coding sequences [29, 30]. This
algorithm exploites a specific way of genetic code degeneration and relations between
mutational pressure and selection pressure shaping the amino acid usage in the pro-
teomes. We used the algorithm to study how selection operating on the nucleotide level
influences the elimination of genes subjected to the directional mutational pressure. We
also analysed changes in the coding potential of genes under the mutational pressure
during the simulation time.

2 Materials and Methods

We decided to perform our simulations on gene sequences of the Borrelia burgdorferi
genome because it shows very strong compositional bias related to differently replicated
DNA strands [14, 15, 31]. In addition to that, it does not show the selection for synony-
mous codon usage, and has the defined mutational pressure associated with replication
for the both DNA strands [3]. Therefore, it appears to be very suitable for studies on
DNA asymmetry and effects of the directional mutational pressure on protein coding
sequences.
The simulation were applied to 475 protein coding sequences extracted from B.
burgdorferi genome as annotated in the NCBI database [32]. This set of gene sequences
will be further called 'genome' or 'individual'. The protein coding sequences that con-
stituted the individual were divided into two subsets:

1. sequences lying on the leading DNA strand (including 333 genes of the total length
356 , 034 nt),
2. sequences lying on the lagging strand (comprising 142 genes of the total length
173 , 796 nt).

We considered in our simulation the population consisting of 72 individuals, which
were eliminated and replaced by others during the simulations run.
A particular Monte Carlo step (MCS) of the simulation comprised two stages:

1. the mutation process of gene sequences,
2. the selection process of individuals.

In the mutation stage, a nucleotide from the genome sequence was substituted by an-
other one. First, it was chosen for mutation using the Poisson process assuming one
mutation per genome by average. Then, the selected nucleotide was substituted by an-
other according to the probability given in one of two substitution matrices for the lead-
ing or the lagging strand, respectively (Tab. 1 and Tab. 2). The matrices describing the
real mutational pressure for the differently replicated DNA strands in the B. burgdorferi
genome were constructed empirically by the comparison of original gene sequences