Difference between revisions of "Team:Paris Bettencourt/Model"

 
(46 intermediate revisions by 5 users not shown)
Line 23: Line 23:
 
     margin-right : 0;
 
     margin-right : 0;
 
     margin-top : 0;
 
     margin-top : 0;
 +
}
 +
 +
#video {
 +
vertical-align : middle;
 +
text-align : center;
 +
padding-top : 50px;
 
}
 
}
  
Line 158: Line 164:
  
  
 +
 +
 +
 +
.titlebox2 {/* Subpanels Title*/
 +
    z-index:1;
 +
    position:relative;
 +
text-align:center;
 +
    background:transparent;
 +
margin-top: -50px;
 +
}
 +
 +
.subpanel3 {
 +
    float: left;
 +
    width:400px;
 +
    height:80px;
 +
    font-size:20px;
 +
    text-align:center;
 +
border-radius:7px;
 +
    margin-top:5px;
 +
    margin-bottom:5px;
 +
    margin-right:5px;
 +
    margin-left:5px;
 +
  }
 +
 +
 +
.subpanel3:hover{
 +
    font-weight:bold;
 +
}
 +
.subpanel3:hover .narrowimg {
 +
    opacity:0.8;
 +
    filter: alpha(opacity=80);
 +
}
  
  
Line 231: Line 269:
 
     <h2 class="red" style="text-align:center;">Goals</h2>
 
     <h2 class="red" style="text-align:center;">Goals</h2>
 
     <ul>
 
     <ul>
         <li>To make a computational model to analyze stain-enzyme dynamics</li>
+
         <li>To better understand the enzyme-stain-fabric system.</li>
         <li>To find optimum parameters</li>
+
         <li>To choose an optimal binding affinity for our fabric binding domain.</li>
 +
<li>To predict how much enzyme activity improvement we can expect.</li>
 
     </ul>
 
     </ul>
 
</div>
 
</div>
Line 245: Line 284:
 
       <h2 class="red" style="text-align:center;">Results</h2>
 
       <h2 class="red" style="text-align:center;">Results</h2>
 
       <ul>
 
       <ul>
             <li> Developed a mass-action model to analyze stain dynamics
+
             <li> Three model classes with diverse assumptions.</li>
             <li> A stochastic computational approach using Gillespie algorithm
+
             <li> Enzyme activity can be optimized at intermediate binding constants near 10<sup>4</sup> M<sup>-1</sup>.</li>
             <li> A diffusion model using explicit finite difference method for three dimensional modeling
+
             <li> We predict a 100 fold improvement from an optimzed enzyme</li>
 
+
 
       </ul>
 
       </ul>
 
</div>
 
</div>
Line 257: Line 295:
 
     <h2 class="red" style="text-align:center;">Methods</h2>
 
     <h2 class="red" style="text-align:center;">Methods</h2>
 
     <ul>
 
     <ul>
         <li>Matlab
+
         <li>MATLAB</li>
         <li>ODE solvers
+
         <li>ODE solvers</li>
         <li>Gillespie Algorithm
+
         <li>Gillespie Algorithm</li>
         <li>Explicit Finite Difference method (FDM)
+
         <li>Explicit Finite Difference method (FDM)</li>
 
     </ul>
 
     </ul>
 
</div>
 
</div>
Line 270: Line 308:
 
<div id="subheader">
 
<div id="subheader">
 
<div id="input">
 
<div id="input">
 +
<div style="margin-top:-50px; margin-bottom:20px; margin-left:60px;">
 +
<div class="panel" >
 +
 +
    <a href="https://2016.igem.org/Team:Paris_Bettencourt/Model#ancre" title="Results">
 +
<div id="thpanel" class="subpanel3">     
 +
  <img class="narrowimg" src="https://static.igem.org/mediawiki/2016/a/a7/Paris_Bettencourt-Result_Button.jpeg" width="400px" height="80px"/>
 +
  <div class="titlebox2">
 +
  Skip to results
 +
  </div>
 +
</div>
 +
      </a>
 +
 +
      <a href="https://2016.igem.org/Team:Paris_Bettencourt/Description" title="Project description">
 +
<div id="thpanel" class="subpanel3">
 +
  <img class="narrowimg" src="https://static.igem.org/mediawiki/2016/2/28/Paris_Bettencourt-Project_Button.jpeg" width="400px" height="80px"/>
 +
  <div class="titlebox2">
 +
    Overview of the project
 +
  </div>
 +
</div>
 +
      </a>
 +
 +
 +
 +
</div>
 +
</div>
 +
 +
 
<h2 class="red">Abstract</h2>
 
<h2 class="red">Abstract</h2>
 
<p>
 
<p>
Line 286: Line 351:
 
<h3> Diffusion Models </h3>
 
<h3> Diffusion Models </h3>
 
Using the Method of Explicit Fine Differences, we represent the system as a three-dimensional grid of discrete volumes. Within a volume, reactions are modelled as Mass Action, but species may diffuse between volumes at the borders. This allows us to account for the fact that different volumes contain different amounts of enzymes. In particular, we allow the enzyme-fabric binding to concentrate the enzyme near the fabric, where it is most effective. But it is important to remember that we model only passive diffusion, not active mixing. So this type of modelling may tend to overstate spatial effects.
 
Using the Method of Explicit Fine Differences, we represent the system as a three-dimensional grid of discrete volumes. Within a volume, reactions are modelled as Mass Action, but species may diffuse between volumes at the borders. This allows us to account for the fact that different volumes contain different amounts of enzymes. In particular, we allow the enzyme-fabric binding to concentrate the enzyme near the fabric, where it is most effective. But it is important to remember that we model only passive diffusion, not active mixing. So this type of modelling may tend to overstate spatial effects.
 
<h3>Challenge: Modelling stain removal in a compact washing machine</h3>
 
 
<p>
 
A typical garment is composed of several square meters of fabric and a typical compact washing machine has a volme of 70 liters.
 
 
<br>
 
 
</p>
 
  
  
Line 301: Line 357:
 
<img src="https://static.igem.org/mediawiki/2016/9/9c/Paris_Bettencourt-Project_Model_Model_Equations.jpg" alt="Quercetin strains degradation" style="width:500px;">
 
<img src="https://static.igem.org/mediawiki/2016/9/9c/Paris_Bettencourt-Project_Model_Model_Equations.jpg" alt="Quercetin strains degradation" style="width:500px;">
 
<p>
 
<p>
<b> Figure X: Schema and reaction equations for an ODE model of stain removal.</b> <b>A</b> An enzyme may reversibly bind to clean fabric or stained fabric. For simplicity we assume that these binding constants are equal. Once the enzyme is bound to stained fabric it may be converted to clean fabric. Stain removal is assumed to be irreversible. <b>B</b> The schema gives rise to two binding equilibria and one irreversible reaction. <b>C</b> Three differential equations capture the dynamics of free enzyme, stain-bound enzyme and clean-fabric-bound enzyme. Not shown are similar equations modeling the unbound stained and clean fabric.
+
<b> Figure 1</b> Schema and reaction equations for an ODE model of stain removal. <b>A</b> An enzyme may reversibly bind to clean fabric or stained fabric. For simplicity we assume that these binding constants are equal. Once the enzyme is bound to stained fabric it may be converted to clean fabric. Stain removal is assumed to be irreversible. <b>B</b> The schema gives rise to two binding equilibria and one irreversible reaction. <b>C</b> Three differential equations capture the dynamics of free enzyme, stain-bound enzyme and clean-fabric-bound enzyme. Not shown are similar equations modeling the unbound stained and clean fabric.
  
 
</p>
 
</p>
Line 309: Line 365:
  
  
<h2 class="red">Results</h2>
+
<h2 class="red" id="ancre">Results</h2>
 +
<div id="figurebox">
 +
<p>
 +
<b>Tabel 1: Key parameters for modeling the enzyme-fabric-stain system. </b> Our model attempts to simulate the case of a white t-shirt stained by a single drop of red wine. The reactions take place within a standard 70 L washing machine. The quantity of enzyme in the system was taken from commercial laundry detergents, which typically contain about 1% enzymes by mass. We take the enzyme CatA as an example, taking its catalysis rate from the BRENDA enzyme database.
 +
</p>
 +
<div style="align:center;">  
  
 +
                        <table style="width:50%; text-align:center; margin: 0 auto;">
 +
                        <tr>
 +
                            <th align="left">Key Parameters</th>
 +
                            <th align="left"> </th>
 +
                        </tr>
 +
                        <tr>
 +
                            <td align="left">Volme of a washing machine</td>
 +
                            <td align="left">70 L</td>
 +
                        </tr>
 +
                        <tr>
 +
                            <td align="left">Volume of a wine stain</td>
 +
                            <td align="left">50 μL</td>
 +
                        </tr>
 +
                        <tr>
 +
                            <td align="left">Malvidin concentration in wine</td>
 +
                            <td align="left">200mg/L</td>
 +
                        </tr>
 +
                        <tr>
 +
                            <td align="left">Area of cotton t-shot</td>
 +
                            <td align="left">150000cm^2</td>
 +
                        </tr>
 +
                        <tr>
 +
                            <td align="left">Weight of cotton</td>
 +
                            <td align="left">20mg/cm^2</td>
 +
                        </tr>
 +
                        <tr>
 +
                            <td align="left">Total mass of detergent</td>
 +
                            <td align="left">5g</td>
 +
                        </tr>
 +
                        <tr>
 +
                            <td align="left">Enzyme fraction in detergent</td>
 +
                            <td align="left">1%</td>
 +
                        </tr>
 +
                        <tr>
 +
                            <td align="left">Activity of CatA enzyme</td>
 +
                            <td align="left">200s^-1</td>
 +
                        </tr>
 +
                        </table>
 +
                        </div>
 +
</div>
 
<h3>Model1: Mass-Action Model</h3>
 
<h3>Model1: Mass-Action Model</h3>
  
Line 317: Line 418:
 
<img src="https://static.igem.org/mediawiki/2016/9/9b/Paris_Bettencourt_Model_Diffusionmodel1.jpg" alt="Massaction_model" style="width:900px;">
 
<img src="https://static.igem.org/mediawiki/2016/9/9b/Paris_Bettencourt_Model_Diffusionmodel1.jpg" alt="Massaction_model" style="width:900px;">
 
<p>
 
<p>
<b> Figure X: Schema and reaction equations for an ODE model of stain removal.</b> <b>A</b> Bla Bla. <b>B</b> Bla Bla
+
<b> Figure 2</b> Computational analysis using Mass-Action model. <b>A</b> The three axis 2D plot consisting of K<sub>D</sub>, Stain fraction and stain remaining after 90mins of enzyme degradation. Enzymatic activity is optimized at a binding activity around 10<sup>-4</sup> M. <b>B</b> Typical parameters that need to be considered for the computational analysis, the optimum K<sub>D</sub> is much lower than typical antibody-antigen interactions (10<sup>-9</sup> M).
 
</p>
 
</p>
 
</div>
 
</div>
Line 326: Line 427:
 
<img src="https://static.igem.org/mediawiki/2016/8/8b/Paris_Bettencourt_Model_Diffusionmodel2.jpg" alt="Massaction_model" style="width:500px;">
 
<img src="https://static.igem.org/mediawiki/2016/8/8b/Paris_Bettencourt_Model_Diffusionmodel2.jpg" alt="Massaction_model" style="width:500px;">
 
<p>
 
<p>
<b> Figure X: Schema and reaction equations for an ODE model of stain removal.</b> <b>A</b> Bla Bla. <b>B</b> Bla Bla
+
<b> Figure 3</b> Stain and bound enzyme percentages estimated using Mass-Action model. <b>A</b> The optimum K<sub>D</sub> for stain removal is found to be 10<sup>-4</sup> when stain percentage is plotted against the time of degradation. <b>B</b> The percentage of different forms of enzyme present in the solution such as free form, enzyme-clean fabric complex, enzyme-dirty fabric complex as a function of K<sub>D</sub> is shown.
 
</p>
 
</p>
 
</div>
 
</div>
Line 332: Line 433:
  
 
<h3>Model2: Stochastic Models </h3>
 
<h3>Model2: Stochastic Models </h3>
 
<h3>Model3: Diffusion Models</h3>
 
 
<div id="figurebox">
 
<div id="figurebox">
 
<div style="text-align:center;">
 
<div style="text-align:center;">
<img src="https://static.igem.org/mediawiki/2016/e/eb/Paris_Bettencourt_Model_diffusiomodel1.jpg" alt="Diffusion_model" style="width:700px;">
+
<video width="600" height="500" controls>
 +
  <source src="https://static.igem.org/mediawiki/2016/5/57/T--Paris_Bettencourt--Gillespie_SSA.mp4">
 +
Your browser does not support the video tag.
 +
</video>
 
<p>
 
<p>
<b> Figure X: Schema and reaction equations for an ODE model of stain removal.</b> <b>A</b> Bla Bla. <b>B</b> Bla Bla
+
<b> Figure 4</b> A Stochastic model of enzyme binding was constructed using near-optimal parameters, as determined by the mass-action model. The K<sub>D</sub> is set to 10<sup>-6</sup> and the stain is assumed to cover a 5% of the fabric. EBc represent senzymes bound to clean fabric; EBd is enzymes bound to dirty fabric and E is free enzymes. Under these conditions, almost all of the enzymes are fabric-bound. Because Gillespie simulations play out with few molecules and small scale, this stain is cleaned much faster than in the mass action model above.
 
</p>
 
</p>
 
</div>
 
</div>
 
</div>
 
</div>
 +
 +
<h3>Model3: Diffusion Models</h3>
  
 
<div id="figurebox">
 
<div id="figurebox">
 
<div style="text-align:center;">
 
<div style="text-align:center;">
<img src="https://static.igem.org/mediawiki/2016/a/a7/Paris_Bettencourt_Model_diffusiomodel2.jpg" alt="Diffusion_model" style="width:700px;">
+
<img style="width: 400px; margin-right:30px;" src="https://static.igem.org/mediawiki/2016/d/d2/Paris_Bettencourt_Model_diffusiomodel3.jpg" alt="Diffusion_model" style="width:500px;">
 
<p>
 
<p>
<b> Figure X: Schema and reaction equations for an ODE model of stain removal.</b> <b>A</b> Bla Bla. <b>B</b> Bla Bla
+
<b>Figure 5</b> Simulations in the diffusion model agree with the mass-action model under most parameters. When diffusion is accounted for, the enzyme activity reaches a maximum near a K<sub>D</sub> of 10<sup>-3</sup>. This is a weaker affinity than predicted by mass action alone, perhaps indicating that diffusion is increasing the effective enzyme concentration.
 
</p>
 
</p>
 
</div>
 
</div>
 
</div>
 
</div>
 +
 +
  
 
<br>
 
<br>
Line 360: Line 466:
 
<h3>Mass-Action Model</h3>
 
<h3>Mass-Action Model</h3>
 
<p>
 
<p>
<br>
+
ODE simulations were performed in MATLAB using the ode23t solver and the indicated parameters.
 
+
 
</p>
 
</p>
  
 
<h3>Stochastic Models </h3>
 
<h3>Stochastic Models </h3>
 
<p>
 
<p>
<br>
+
Gillespie's algorithm was performed numerically in MATLAB. Parameters from the mass action model were scaled down to accomodate a countable number of single molecules. The cloth is modeled as grid of binding sites 100 units square, of which 5% are marked as stained. 100 free enzymes are introduced to the system at time zero and the reaction continutes until all the stains are removed.
 
+
 
</p>
 
</p>
  
 
<h3>Diffusion Models</h3>
 
<h3>Diffusion Models</h3>
 
<p>
 
<p>
Explicit Finite Different scheme is used to model three dimensional stain - enzyme dynamics. The enzyme is assumed to be homogeneously spread through out the spatial domain at the start of the experiment. The scheme was applied on a reaction and diffusion equation thereafter. No flux boundary condition was applied at all boundaries which specifically meant for zero enzyme loss from the system. One of the boundaries is taken as the shirt with stain (1cm^2 area). The parameters and the initial conditions used in the simulations were chosen as realistic as possible. MATLAB was used to computationally model the system and perform thee simulations.
+
The Explicit Finite Difference Method was perfomed numerically in MATLAB. The enzyme was initially assumed to be homogenously spread in the spatial domain. A no flux boundary condition was applied, preventing enzymes from leaving the system. One of the boundaries is simulated as a shirt of 1 m<sup>2</sup> ares with a 1 cm<sup>2</sup>. Other parameters were chosen as in the mass action model.
 
<br>
 
<br>
  
Line 382: Line 486:
 
<h2 class="red">Attributions</h3>
 
<h2 class="red">Attributions</h3>
 
<p>
 
<p>
This project was done mostly by Jake Wintermute, Mislav Acman and Mani Sai Suryateja Jammalamadaka.   
+
This project was done mostly by Mislav Acman and Mani Sai Suryateja Jammalamadaka.   
 
</p>
 
</p>
 
+
<div style="text-align:center;"> <img src="https://static.igem.org/mediawiki/2016/e/ee/Paris_Bettencourt-tejastatic.jpeg" width="200px" class="img-rounded"/><img src="https://static.igem.org/mediawiki/2016/4/4e/Paris_Bettencourt-mislavstatic.jpeg" width="200px" class="img-rounded"/></div>
 
<h2 class="red">References</h2>
 
<h2 class="red">References</h2>
 
<ul>
 
<ul>
 
         <li> Enzyme Database- BRENDA
 
         <li> Enzyme Database- BRENDA
 
         <li> Numerical Analysis and Optimization, An Introduction to Mathematical Modelling and Numerical Simulation- Grégoire Allaire
 
         <li> Numerical Analysis and Optimization, An Introduction to Mathematical Modelling and Numerical Simulation- Grégoire Allaire
        <li>
 
 
</ul>
 
</ul>
  

Latest revision as of 03:44, 20 October 2016



Goals

  • To better understand the enzyme-stain-fabric system.
  • To choose an optimal binding affinity for our fabric binding domain.
  • To predict how much enzyme activity improvement we can expect.

Results

  • Three model classes with diverse assumptions.
  • Enzyme activity can be optimized at intermediate binding constants near 104 M-1.
  • We predict a 100 fold improvement from an optimzed enzyme

Methods

  • MATLAB
  • ODE solvers
  • Gillespie Algorithm
  • Explicit Finite Difference method (FDM)

Abstract

To act on stains, an enzyme must be concentrated at the fabric surface. Our project began with the idea that we can effectively increase this concentration with a fabric binding domain (FBD). But does this idea hold up to detailed scrutiny? What is the optimal affinity for a stain removing enzyme? How much activity improvement can we expect to achieve? To answer these questions, we built three models of the enzyme-fabric-stain interaction: a differential equation model, a stochastic process model and a 3D reaction-diffusion model. The results of these models agreed on three main points. At low affinity, enzymes diffuse away from fabric and into solution. At high affinity, enzymes become trapped on clean sections of fabric and are unable to reach the stain. But for a wide range of conditions, enzymatic activity is optimized at a binding activity around 10-4 M. For reference, this affinity is much lower than typical antibody-antigen interactions (10-9 M) and is within reach of our protein engineering methods. Under realistic conditions we predict that an optimal binding domain will improve activity by 100 or 1000 fold.

Motivation and Background

Mass-Action Models

In the simplest model, reactant concentrations are assumed to be uniform throughout the system. We make the mass action assumption: that reaction rates are directly proportional to reactant concentrations. In this way, we can represent the system with just a few simple differential equations. These great advantage of these models is conceptual simplicity and computational speed. It is possible to repeat the simulations many times over large parameter ranges. But it is important to remember that these models do not account for spatial structure. It is as though the fabric has been split into many small pieces and thoroughly mixed.

Stochastic Models

With Gillespie simulations, we are able to keep track of individual molecules. This allows us to account for single molecule dynamics that aren't part of mass action modes. In a mass action model, interactions are averaged over large numbers and continuous, but in a Gillespie model they are discrete.For example, a single enzyme may bind and unbind a stain without ever acting on a stain. A disadvantage of Gillespie simulations is that they can only by computed for very small numbers of molecules. They also do not represent spatial structure.

Diffusion Models

Using the Method of Explicit Fine Differences, we represent the system as a three-dimensional grid of discrete volumes. Within a volume, reactions are modelled as Mass Action, but species may diffuse between volumes at the borders. This allows us to account for the fact that different volumes contain different amounts of enzymes. In particular, we allow the enzyme-fabric binding to concentrate the enzyme near the fabric, where it is most effective. But it is important to remember that we model only passive diffusion, not active mixing. So this type of modelling may tend to overstate spatial effects.
Quercetin strains degradation

Figure 1 Schema and reaction equations for an ODE model of stain removal. A An enzyme may reversibly bind to clean fabric or stained fabric. For simplicity we assume that these binding constants are equal. Once the enzyme is bound to stained fabric it may be converted to clean fabric. Stain removal is assumed to be irreversible. B The schema gives rise to two binding equilibria and one irreversible reaction. C Three differential equations capture the dynamics of free enzyme, stain-bound enzyme and clean-fabric-bound enzyme. Not shown are similar equations modeling the unbound stained and clean fabric.

Results

Tabel 1: Key parameters for modeling the enzyme-fabric-stain system. Our model attempts to simulate the case of a white t-shirt stained by a single drop of red wine. The reactions take place within a standard 70 L washing machine. The quantity of enzyme in the system was taken from commercial laundry detergents, which typically contain about 1% enzymes by mass. We take the enzyme CatA as an example, taking its catalysis rate from the BRENDA enzyme database.

Key Parameters
Volme of a washing machine 70 L
Volume of a wine stain 50 μL
Malvidin concentration in wine 200mg/L
Area of cotton t-shot 150000cm^2
Weight of cotton 20mg/cm^2
Total mass of detergent 5g
Enzyme fraction in detergent 1%
Activity of CatA enzyme 200s^-1

Model1: Mass-Action Model

Massaction_model

Figure 2 Computational analysis using Mass-Action model. A The three axis 2D plot consisting of KD, Stain fraction and stain remaining after 90mins of enzyme degradation. Enzymatic activity is optimized at a binding activity around 10-4 M. B Typical parameters that need to be considered for the computational analysis, the optimum KD is much lower than typical antibody-antigen interactions (10-9 M).

Massaction_model

Figure 3 Stain and bound enzyme percentages estimated using Mass-Action model. A The optimum KD for stain removal is found to be 10-4 when stain percentage is plotted against the time of degradation. B The percentage of different forms of enzyme present in the solution such as free form, enzyme-clean fabric complex, enzyme-dirty fabric complex as a function of KD is shown.

Model2: Stochastic Models

Figure 4 A Stochastic model of enzyme binding was constructed using near-optimal parameters, as determined by the mass-action model. The KD is set to 10-6 and the stain is assumed to cover a 5% of the fabric. EBc represent senzymes bound to clean fabric; EBd is enzymes bound to dirty fabric and E is free enzymes. Under these conditions, almost all of the enzymes are fabric-bound. Because Gillespie simulations play out with few molecules and small scale, this stain is cleaned much faster than in the mass action model above.

Model3: Diffusion Models

Diffusion_model

Figure 5 Simulations in the diffusion model agree with the mass-action model under most parameters. When diffusion is accounted for, the enzyme activity reaches a maximum near a KD of 10-3. This is a weaker affinity than predicted by mass action alone, perhaps indicating that diffusion is increasing the effective enzyme concentration.


Methods

Mass-Action Model

ODE simulations were performed in MATLAB using the ode23t solver and the indicated parameters.

Stochastic Models

Gillespie's algorithm was performed numerically in MATLAB. Parameters from the mass action model were scaled down to accomodate a countable number of single molecules. The cloth is modeled as grid of binding sites 100 units square, of which 5% are marked as stained. 100 free enzymes are introduced to the system at time zero and the reaction continutes until all the stains are removed.

Diffusion Models

The Explicit Finite Difference Method was perfomed numerically in MATLAB. The enzyme was initially assumed to be homogenously spread in the spatial domain. A no flux boundary condition was applied, preventing enzymes from leaving the system. One of the boundaries is simulated as a shirt of 1 m2 ares with a 1 cm2. Other parameters were chosen as in the mass action model.

Attributions

This project was done mostly by Mislav Acman and Mani Sai Suryateja Jammalamadaka.

References

  • Enzyme Database- BRENDA
  • Numerical Analysis and Optimization, An Introduction to Mathematical Modelling and Numerical Simulation- Grégoire Allaire
Centre for Research and Interdisciplinarity (CRI)
Faculty of Medicine Cochin Port-Royal, South wing, 2nd floor
Paris Descartes University
24, rue du Faubourg Saint Jacques
75014 Paris, France
+33 1 44 41 25 22/25
igem2016parisbettencourt@gmail.com
2016.igem.org