Sie sind auf Seite 1von 3

Introduction to Computational Biology 03-250/02-250/03-252/02-252, Spring 2016

R. F. Murphy

Recitation 11 Solutions
Deterministic and Stochastic Gene Regulation Models
The information encoded in DNA sequences of genes is expressed in the form of proteins through
multiple biochemical steps. The amount of protein produced from a given gene depends on the
following processes:
1) Gene activation and inactivation.
2) Transcription and turnover of mRNA.
3) Translation and turnover of proteins.
In this recitation, you will develop models of gene expression that take into account each of these
processes. You are given Matlab and Python code to start from. You will begin with a deterministic
model based on a set of three coupled ordinary differential equations (one for each of the above
processes) and investigate the effects of varying parameters involved in these processes.
Deterministic model
1. Given the following reaction diagram, write a function to numerically solve the following set of
ODEs for gene expression:
a.

dn1
= 1+ ( n1max n1 ) 1 * n1 (Gene activation / inactivation)
dt

b.

dn 2
= 2 * n1 n 2 / 2
dt

(Transcription and turnover of mRNA)

c.

dn 3
= 3 * n 2 n 3 / 3
dt

(Translation and turnover of proteins)

Where n1, n2, and n3 represent the number of active genes, mRNA transcripts, and protein
molecules, respectively. 1+ and 1 represent the gene activation and inactivation rates. 2 and

3 give the transcription and translation rates, and 2 and 3 give the average lifetime of mRNA
and protein molecules. Use the following parameters and initial values:

activegenes)
n1 (number of
n2 (number of mRNA)
n3 (number of Protein)
1+ (gene activation rate)
1 (gene inactivation rate)
2 (transcription rate)
3 (translation rate)
2 (RNA lifetime)
3 (protein lifetime)

Page 1 of 3

2 (initial)
0 (initial)
0 (initial)
0.5
1
1
0.1
20
100

Introduction to Computational Biology 03-250/02-250/03-252/02-252, Spring 2016

R. F. Murphy

Solution: See code.


2. For these parameters, what are the steady-state values for the number of:
a. Active gene copies?
0.66
b. mRNA transcripts?
13.33
c. Protein molecules?
133
3. In what order do the species (gene, mRNA, protein) reach steady state? Does this make sense?
Why or why not?
Active gene copies, then mRNA, then protein. This makes sense, as mRNA levels will
continue to change until active gene copies have stabilized, and protein levels will
continue to change until mRNA levels reach steady-state.
4. Lets see what happens to the model when we alter the rates of transcription and translation.
Switch the rates of transcription and translation, such that transcription is slow (0.1) and
translation is fast (1) and rerun the model. What happens to mRNA and protein levels in the
deterministic model?
With a 10-fold lower transcription rate, the steady-state level of mRNA decreases 10fold. The 10-fold increase in translation rates compensate for the reduced mRNA level,
leaving the steady-state protein level unchanged.
Stochastic model
In individual cells, mRNA molecules transcribed from individual genes are often rare. For example, the
median mRNA copy numbers per cell in E. coli, bakers yeast and mouse fibroblasts are 0.25, 1, and 17.
As a result, stochastic models provide more accurate representations of gene expression. Next we will
use an implementation of Gillespies algorithm to stochastically simulate the same process described in
the ODEs above.
Open GillespieSim.m. This program uses a while loop to implement the Gillespie algorithm.
5. Read the program and comments, and answer the following questions:
a. Where are the rate constants stored?
In the c vector.
b. What do the values in the changeVector represent? Describe what the first row means in
terms of changes to the state variables of Gene, RNA, and Protein.
The changeVector is a matrix of vectors used to update the state of the system
after an event occurs. The first row increases the Gene variable by 1 and doesnt
change the mRNA or Protein variables, reflecting with the occurrence a gene
activation event.
c. Consider the alpha vector. How do the variables correspond to the ODEs in the
deterministic model?
The first term, c(1)*(n1max-x(1)), is equivalent to the gene activation expression
1+ ( n1max n1 ) . The second corresponds with gene inactivation, etc. These give the
propensities of each event type.

Page 2 of 3

Introduction to Computational Biology 03-250/02-250/03-252/02-252, Spring 2016

R. F. Murphy

6. Run the simulation for 500 minutes with GillespieSim(500) a few times. Whats the average
protein level during the run? How does this compare with the steady-state protein level of the
deterministic model?
Running [X,Y] = GillespieSim(500) stores the gene, mRNA and protein count data in the
Y matrix. mean(Y(:,3)) gives the mean protein value. Its lower than the deterministic
model on this time scale. If you evaluate the mean for timepoints 200-500 (with
mean(Y(200:end, 3))), its generally close to the deterministic steady-state value.
7. As with the deterministic model evaluate the consequences of switching the rates of transcription
and translation. Evaluate the model with the transrate parameter set to 0.1 and the translatrate
parameter set to 1. How does this affect the shape or noisiness of the plot?
As transcription rates decrease, mRNA levels remain lower and fluctuate more rapidly.
Protein levels also fluctuate more rapidly. This makes the protein levels appear more
noisy.
2
8. We can quantify noisiness as the ratio of the variance to the mean
, also known as the index

of dispersion (IOD). The matlab program NoiseLoop.m calculates this measure of noise by
running the Gillespie simulation 50 times and storing the IOD for each simulation run. It then
returns the median IOD for the 50 runs and plots a histogram of IOD frequencies.
translation rates to those in the table
a. Reset the GillespieSim.m program transcription and
on page 1. Run the NoiseLoop a few times and record the median IOD.
The median IOD is ~ 1.8.
b. Switch the transcription and translation rates in the GillespieSim.m program and run the
NoiseLoop a few times to record the median IOD.
The median IOD is ~ 8.
9. Do the results of the model agree with your visual observations from question 8? How can you
explain your results?
The simulations show that fast rates of transcription and slow rates of translation result in
reduced noise compared to the opposite (slow transcription and fast translation). This
agrees with the visual observations of the model results. I can explain the results because
slow translation rates make protein levels less responsive to fluctuations in gene
activation and mRNA abundance.

Page 3 of 3

Das könnte Ihnen auch gefallen