Team:ETH Zurich/Modeling/Genome Static Analysis

Restriction Enzyme Analysis

This section presents the computational investigation we performed in order to understand which restriction enzymes are optimal when used in our minimal genome approach, in order to find out which restriction enzymes cut the genome in fragments that most probably will lead to find the minimal genome in our reduction approach. Important is to note that this is a "statical" analysis, meaning that we do not include in the evaluation of the restriction enzyme optimality any prevision regarding to the effects its cutting patterns can have on cell physiology or cell system behaviour. We addressed questions regarding the cell system response after genome reduction using more advanced modelling techniques (a genome scale model) in the Genome Scale Analysis section. We focus here only on the insights that can be obtained using three kinds of "statical" information:

• the genome sequence of our strain of interest (E.Coli K12 MG1655).
• the annotation information of our strain of interest (E.Coli K12 MG1655).
• the recognition site patterns of each of the restriction enzymes we test.

Using computational tools and the above mentioned information we are interested in asking (and answering) the following questions:

• Which are the available restriction enzymes, their recognition sites and the fragments they generate after digestion?
• How is the distribution of the genes in each fragment related to the frequence of cutting?
• Is it possible to identify restriction enzymes that optimizes the probability of cutting out fragments of the genome but still keeping the cell alive (or better, do exist restriction enzymes that rearely targets fragment containing essential genes)?

Available restricion enzymes and digestion simulation

As source for the restriction enzyme to consider, we used the [http://rebase.neb.com/rebase/rebase.html REBASE database]. We found 713 restriction enzymes that spawn from 4 up to 13 cutters, some with complete specific recognition sites and some with unspeficisity properties. Since some of the restriction enzymes present the same recognition site sequence, we grouped them together as a single entity to be tested (216 groups). We downloaded the genome and annotation information regarding E.Coli K12 MG1655 from GenBank® database. We then simulated the digestion of E.Coli chromosome sequentially for each group of restriction enzymes and performed statistical analysis on the fragment pattern obtained. The following pictures summarize the distribution of the available enzymes regarding to their frequency of cutting (number of fragments after digestion):

It is possible to note that there is a huge number of restriction enzymes that digest the chromosome in few to high number of fragments (up to 10000 fragments) and relatively fewer that generate a very high number of fragments.

Analysing the gene content of the fragments

In order to understand if there are restriction enzymes that have particular properties (for example the ability to target on the same fragments several essential genes, in order to reduce the probability of causing cell death) we performed some statistical analysis, calculating indexes such as: the mean and variance for fragment lengths, the mean and variance for gene numbers per fragment, the probability of one fragment to contain an essential gene. Here above we show the graphs obtained by plotting these indexes.

identified in the devised mechanism. Accordingly, we divided the modelling framework in four modules that tack the relative problematics.
The first module is concerned with the analysis of restriction enzymes and their cutting pattern on E.Coli genome, the second module predicts the cell's response to the selection pressure and the forced genome reduction from a system point of view (that is, using a genome scale model), the third module addresses issues related to the sensitivity and setting of the chemostat mechanism, the fourth and final module presents the model of the genetic switch circuit used to control the restriction enzymes expression.
In the table below, you can find a bird-eye view on the four modules, with the most important aspects highlighted. Since we believe that a model is useful only when it answers specific and well-posed questions, this is the first aspect we report in the summary view. Second we briefly report about the modelling method applied. As last, we summarize the results we obtained.
By clicking on each module's title, you can browse the specific module pages containing all the detailed information, such as plots, modelling assumptions and data sources. Note that is possible to download from our Downloading Page all the data and code (MATLAB source) that we used and wrote in order to generate our results.

Obviously the mean number fo genes per fragment follows a linear relation with the number of fragments in a log-log plot. Used as a check for the validity of our fragmentation algorithm.

The variance of the number of genes per fragment is very well correlated to the number of fragments.

The variance in fragments size is linearly correlated in a log-log plot with the variance in genes per fragment.

The probability of each fragment to contain an essential gene gives to us a criteria for understanding the level of mortality of a particular fragmentation pattern. This relation also is determined by the frequency of cutting.

As conclusion, we can state that from the static analysis is not possible to discriminate optimal restriction enzymes. It is evident that known (essential) genes on the chromosome are randomly distributed, as well as the cutting sites of restriction enzymes. Our choice of the restriction enzyme to be used should then based only one the frequency of cutting and related issues, such as the efficiency of cutting, and on the genome scale model results.

Result table on chromosomal digestion simulations