A Whole-Cell Computational Model Predicts Phenotype from Genotype - - PowerPoint PPT Presentation

A Whole-Cell Computational Model Predicts Phenotype from Genotype

Computational Models in Biology

• Mathematically model cellular pathways to

predict phenotypes

• Quantify known pieces of data to analyze the

implementation of information in a cell

• More cost and time effective when compared

to phenotypic assays

Mycoplasma genitalium

• Small parasitic bacterium

○ Mycoplasma ○ Evolved through genome reduction ○ Lack cell wall

• Smallest genome known

○ 580 kb ○ Self-replicating

Mycoplasma genitalium

• Discovered during drug therapy studies in

1970s

○ Investigating acute nongonococcal urethritis (NGU)

• Replicates through binary fission
• Adheres to glass and plastic surfaces

○ In the body: epithelial lining, respiratory pathways, spermatozoa, and erythrocytes

Pathogenesis

• Toxin: MG-186

○ Degrades host nucleic acids to use for its own growth

• Localizes in immune

cells

○ Prevents complete killing of bacteria ○ Low level infection persists

• Human urogential

tract is preferred site for colonization

Disease

• Considered an STI and can be passed

through unprotected sex

• Acute nongonococcal urethritis (NGU)

○ Inflammation of the urethra (males)

• Symptoms

○ Burning during urination ○ Discharge ○ Infertility ■ Adhere to spermatozoa

• Difficult to diagnose

○ Usually a secondary infection

Treatment

• 1 week time course of antibiotics is

recommended by the CDC

○ Azithromycin ○ Doxycycline ○ Erythromycin

• A vaccine has not been developed to

prevent M. genitalium infection

Whole-cell

• Common metabolites
• DNA
• RNA
• Protein
• and “Other”

28 independent submodels fit within these five categories of variables

Diagram of the whole-cell model of Mycoplasma genitalium

• The authors started out with the assumption that each submodel within the cell

was approximately independent from every other submodel on timescales of <1s.

• Simulations of the cell processes were performed as running through a loop,

where each submodel was running independently from each other but was dependent on values from other submodels from the previous time step. 28 independent sub-models 16 cell variables

• C. Comparison of predicted percent dry mass of the cell,

broken into four components, with experimental data. Blue is predicted mass, Black is experimental data from Morowitz et al. (1962).

• A and B. Comparison of experimental (A) and predicted

(B) doubling times. Model validation against experimental data

Model validation against experimental data

• G. Simulated data from
• ne cell showing the

correlation between mRNA synthesis (HMW2 mRNA) and “bursts” of protein synthesis

• H. The lack of correlation

between mRNA and protein counts in 128 individual cells is congruent with previous single-cell experiments by Taniquchi et al (2010).

So, the authors’ modeling again seems to be validated by experimental data.

Now they can do more interesting simulations:

For E and F, they simulated 128 individual cells throughout the cell cycle for the binding and collision frequencies of DNA-binding proteins.

• E. Predicted frequency of binding and displacement of

DNA-binding proteins.

• F. The correlation of protein density and collisions over

time.

• B. Within 20 minutes from the

start of the cell cycle, 90% of the chromosome has been traversed by one protein or another (most likely RNA-pol).

• C. On average, 90% of the genes

have been expressed within the first 2.5 hours.

More simulations on DNA-proteins interactions

Transcription and translation dominated energy usage (by percentage of ATP and GTP used). Clearly synchronous processes

• Additionally, the authors carried out simulations for quantitative

characterization gene disruption – its effect on growth and other parameters...

Δ Δ

• Model is represented as a directed graph.
• At each node there is conservation of flow:

∑ Inflows=∑ Outflows

• Vary the flows to:
• Optimize Growth
• Subject to constraints