Team:ETHZ Basel/Modeling/Experimental Design

From 2010.igem.org

(Difference between revisions)
(Evaluator)
 
(23 intermediate revisions not shown)
Line 2: Line 2:
{{ETHZ_Basel10_Modeling}}
{{ETHZ_Basel10_Modeling}}
-
= Interworking: From mathematical modeling to biology and back =
+
= Experimental Design =
-
E. lemming is a special project, because the final product has biological and computer science aspects in equal parts. For this reason, it was decided to connect the wet laboratory and the modeling subteam closely from the very beginning to make the best decisions for the whole project. Interworking is the process, where two parts interact on each other to create a comprehensive result. This section describes the insight gained by mathematical modeling for both wet laboratory and information processing.
+
Since the design of E.lemming implies the existence of many possible biological combinations of the fundamental parts, the first steps in the modeling of our system were directed towards reducing the actual number of combinations implemented in the wetlab. Furthermore, by using the combined molecular models for ''in silico'' evaluation of the best possible devices, we derived theoretical results on choosing the biological parts which maximize the chance of a functional final ensemble.
-
== Insights for wet laboratory ==
+
[[Team:ETHZ_Basel/Modeling/Experimental_Design#Insights_for_wet_laboratory|''' Wet laboratory evaluation results''']] have showed that molecular modeling and experimental biology can interwork to gain new insight on both aspects of our project.  
-
[[Image:ETHZ_Basel_molecular_eval.png|thumb|400px|'''Models used for evaluation.''' Coupled individual models for the simulation of the whole process and their interfaces.]]
+
 +
[[Image:ETHZ_Basel_molecular_eval.png|thumb|400px|'''Models used for experimental design.''' Coupled individual models for the simulation of the whole process and their interfaces.]]
 +
 +
== Insights for wet laboratory ==
=== Goals of the evaluation ===
=== Goals of the evaluation ===
-
To determine the best possible parts for the biological implementation of E. lemming, the combined molecular model (Light switch - Chemotaxis) has been used to answer specific questions:
+
To determine the best network structure for the biological implementation of the E. lemming and the best experimental conditions, the combined molecular model (PhyB/PIF3 light switch - Chemotaxis) has been used to answer specific questions:
# Which protein of the chemotaxis pathway (CheR, CheB, CheY or CheZ) should be fused...
# Which protein of the chemotaxis pathway (CheR, CheB, CheY or CheZ) should be fused...
Line 29: Line 31:
! Che !! LSP1 !! LSP2 !! [Asp] !! [AP] !! [anchor] !! model
! Che !! LSP1 !! LSP2 !! [Asp] !! [AP] !! [anchor] !! model
|-
|-
-
| CheR || PhyB || PIF3 || 0 uM || 0 uM || 0 uM || Spiro
+
| CheR || PhyB || PIF3 || 0 μM || 0 μM || 0 μM || Spiro
|-
|-
-
| CheB || PIF3 || PhyB || 10^-6 uM || 25 uM || 25 uM || Mello
+
| CheB || PIF3 || PhyB || 10<sup>-6</sup> μM || 25 μM || 25 μM || Mello
|-
|-
-
| CheY |||||| 10^-3 uM || 50 uM || 50 uM ||
+
| CheY |||||| 10<sup>-3</sup> μM || 50 μM || 50 μM ||
|-
|-
-
| CheZ |||||||| 75 uM || 75 uM ||
+
| CheZ |||||||| 75 μM || 75 μM ||
|-
|-
-
||||||||| 100 uM || 100 uM ||
+
||||||||| 100 μM || 100 μM ||
|}
|}
Line 46: Line 48:
{| border="0" align="center"
{| border="0" align="center"
|- valign="top"
|- valign="top"
-
|[[Image:ETHZ_Basel_models_evaluator_cher.png|thumb|200px|'''Che = CheR.'''<br><br>Amplitude :0.25<br>(CheR, LSP1:PIF3, Asp:Medium, Model:Spiro)]]
+
|[[Image:ETHZ_Basel_models_evaluator_cher.png|thumb|200px|'''Che = CheR.'''<br><br>Amplitude: 0.35<br>(CheR, LSP1:PIF3, Asp:Medium, Model:Spiro)]]
-
|[[Image:ETHZ_Basel_models_evaluator_cheb.png|thumb|200px|'''Che = CheB.''' Inverse relation of activation / inactivation of CheB limit possible applications in E. lemming.<br><br>Amplitude: 0.16<br>(CheB, LSP1:PIF3, Asp:Medium, Model:Spiro)]]
+
|[[Image:ETHZ_Basel_models_evaluator_cheb.png|thumb|200px|'''Che = CheB.''' Inverse relation of activation / inactivation of CheB limit possible applications in E. lemming.<br><br>Amplitude: 0.35<br>(CheB, LSP1:PIF3, Asp:Medium, Model:Spiro)]]
-
|[[Image:ETHZ_Basel_models_evaluator_chey.png|thumb|200px|'''Che = CheY.'''<br><br>Amplitude: 0.8<br>(CheY, LSP1:PIF3, Asp:Medium, Model:Spiro)]]
+
|[[Image:ETHZ_Basel_models_evaluator_chey.png|thumb|200px|'''Che = CheY.'''<br><br>Amplitude: 0.9<br>(CheY, LSP1:PIF3, Asp:Medium, Model:Spiro)]]
-
|[[Image:ETHZ_Basel_models_evaluator_chez.png|thumb|200px|'''Che = CheZ.''' CheZ showed the most increase in terms of relative amplitude.<br><br>Amplitude: 1.4<br>(CheZ, LSP1:PIF3, Asp:Medium, Model:Spiro)]]
+
|[[Image:ETHZ_Basel_models_evaluator_chez.png|thumb|200px|'''Che = CheZ.''' CheZ showed the most increase in terms of relative amplitude.<br><br>Amplitude: 1.6<br>(CheZ, LSP1:PIF3, Asp:Medium, Model:Spiro)]]
|}
|}
Line 59: Line 61:
{| border="0" align="center"
{| border="0" align="center"
|- valign="top"
|- valign="top"
-
|[[Image:ETHZ_Basel_models_evaluator_phyb.png|thumb|400px|'''LSP1 = PhyB.''' PhyB coupling is very robust and results in maximal plateau surface.<br><br>Amplitude: 0.3<br>(CheR, LSP1:PhyB, Asp:Medium, Model:Spiro)]]
+
|[[Image:ETHZ_Basel_models_evaluator_phyb.png|thumb|400px|'''LSP1 = PhyB.''' PhyB coupling is very robust and results in maximal plateau surface.<br><br>Amplitude: 0.5<br>(CheY, LSP1:PhyB, Asp:Medium, Model:Spiro)]]
-
|[[Image:ETHZ_Basel_models_evaluator_pif3.png|thumb|400px|'''LSP1 = PIF3.''' Coupling to PIF3 results in a much more sensitive area of [AP] vs. [anchor].<br><br>Amplitude: 0.25<br>(CheR, LSP1:PIF3, Asp:Medium, Model:Spiro)]]
+
|[[Image:ETHZ_Basel_models_evaluator_pif3.png|thumb|400px|'''LSP1 = PIF3.''' Coupling to PIF3 results in a much more sensitive area of [AP] vs. [anchor].<br><br>Amplitude: 0.9<br>(CheY, LSP1:PIF3, Asp:Medium, Model:Spiro)]]
|}
|}
-
PhyB is according to the evaluation the more robust choice compared to PIF3, but it's speculated that PhyB might have the tendency to sequester, since PhyA does under certain conditions. This would result in a lowered active concentration of the species, because the proteins stick together. In addition, PIF3 is much smaller than PhyB and it could therefore have a lower negative influence on the Che protein. Relative amplitude of both combinations is very similar. For this reasons, PIF3 was the first choice as light-sensitive protein linked to the Che protein.
+
PhyB is according to the evaluation the more robust choice compared to PIF3, but it's speculated that PhyB might have the tendency to sequester, since PhyA does under certain conditions. This would result in a lowered active concentration of the species, because the proteins stick together. In addition, PIF3 is much smaller than PhyB and it could therefore have a lower negative influence on the Che protein. Relative amplitude of PIF3 is higher, although at lower levels, it's very similar. For this reasons, PIF3 was the first choice as light-sensitive protein linked to the Che protein.
==== 3. What is a good aspartate concentration? ====
==== 3. What is a good aspartate concentration? ====
{| border="0" align="center"
{| border="0" align="center"
|- valign="top"
|- valign="top"
-
|[[Image:ETHZ_Basel_models_evaluator_asp0.png|thumb|260px|'''[Asp] = 0 uM.''' No aspartate in the systems results in very low relative amplitude.<br><br>Amplitude: 0.018<br>(CheR, LSP1:PIF3, Asp:Zero, Model:Spiro)]]
+
|[[Image:ETHZ_Basel_models_evaluator_asp0.png|thumb|260px|'''[Asp] = 0 μM.''' No aspartate in the systems results in very low relative amplitude.<br><br>Amplitude: 0.03<br>(CheR, LSP1:PIF3, Asp:Zero, Model:Spiro)]]
-
|[[Image:ETHZ_Basel_models_evaluator_asp1.png|thumb|260px|'''[Asp] = 10^-6 uM.''' Medium aspartate levels show reasonable amplitude levels of amplitude.<br><br>Amplitude: 0.25<br>(CheR, LSP1:PIF3, Asp:Medium, Model:Spiro)]]
+
|[[Image:ETHZ_Basel_models_evaluator_asp1.png|thumb|260px|'''[Asp] = 10^-6 μM.''' Medium aspartate levels show reasonable amplitude levels of amplitude.<br><br>Amplitude: 0.35<br>(CheR, LSP1:PIF3, Asp:Medium, Model:Spiro)]]
-
|[[Image:ETHZ_Basel_models_evaluator_asp2.png|thumb|260px|'''[Asp] = 10^-3 uM.''' High aspartate levels reach saturation of receptors and cannot increase the relative amplitude much further.<br><br>Amplitude: 0.35<br>(CheR, LSP1:PIF3, Asp:High, Model:Spiro)]]
+
|[[Image:ETHZ_Basel_models_evaluator_asp2.png|thumb|260px|'''[Asp] = 10^-3 μM.''' High aspartate levels reach saturation of receptors and cannot increase the relative amplitude much further.<br><br>Amplitude: 0.5<br>(CheR, LSP1:PIF3, Asp:High, Model:Spiro)]]
|}
|}
Line 78: Line 80:
{| border="0" align="center"
{| border="0" align="center"
|- valign="top"
|- valign="top"
-
|[[Image:ETHZ_Basel_models_evaluator_spiro.png|thumb|400px|'''Spiro prediction with first choice parameters.'''<br><br>Amplitude: 0.8<br>(CheY, LSP1:PIF3, Asp:Medium, Model:Spiro)]]
+
|[[Image:ETHZ_Basel_models_evaluator_spiro.png|thumb|400px|'''Spiro prediction with first choice parameters.'''<br><br>Amplitude: 0.9<br>(CheY, LSP1:PIF3, Asp:Medium, Model:Spiro)]]
-
|[[Image:ETHZ_Basel_models_evaluator_mello.png|thumb|400px|'''Mello prediction with first choice parameters.'''<br><br>Amplitude: 0.8<br>(CheY, LSP1:PIF3, Asp:Medium, Model:Mello)]]
+
|[[Image:ETHZ_Basel_models_evaluator_mello.png|thumb|400px|'''Mello prediction with first choice parameters.'''<br><br>Amplitude: 0.9<br>(CheY, LSP1:PIF3, Asp:Medium, Model:Mello)]]
|}
|}
-
According to the the evaluation of all 48 parameter combination possibilities, it was decided to take a concentration for anchor plasmid of [anchor] = 50 uM and a concentration for the anchor binding protein [AP] = 40 - 50 uM, which is slightly below to ensure having enough binding positions.
+
According to the the evaluation of all 48 parameter combination possibilities, it was decided to take a concentration for anchor plasmid of [anchor] = 50 μM and a concentration for the anchor binding protein [AP] = 40 - 50 μM, which is slightly below to ensure having enough binding positions.
==== Comparison of the chemotaxis models ====
==== Comparison of the chemotaxis models ====
Line 92: Line 94:
=== Evaluator ===
=== Evaluator ===
-
Again, the output variable was the concentration of free and phosphorylated CheYp. To determine the time constants, the time elapsed upon activation of the system until the amplitude reached 2^-(1/2) of the local maximum was measured. Results were determined for our target configuration of E. lemming (CheY, LSP1:PIF3, Asp:Medium).
+
Again, the output variable was the concentration of free and phosphorylated CheYp. To determine the time constants, the time elapsed upon activation of the system until the amplitude reached 2<sup>-1/2</sup> of the local maximum was measured. Results were determined for our target configuration of E. lemming (CheY, LSP1:PIF3, Asp:Medium).
=== Results ===
=== Results ===
{| border="0" align="center"
{| border="0" align="center"
|- valign="top"
|- valign="top"
-
|[[Image:ETHZ_Basel_models_evaluator_ip_spiro_frl.png|thumb|400px|'''Evaluation of activation (FRL) time constant as predicted by the model from Spiro et al. (1997): 0.3485''' Time elapsed upon activation of the system until the amplitude reached 2^-(1/2) of the local maximum was measured.]]
+
|[[Image:ETHZ_Basel_models_evaluator_ip_spiro_frl.png|thumb|400px|'''Evaluation of activation (FRL) time constant as predicted by the model from Spiro et al. (1997): 0.4092s''' Time elapsed upon activation of the system until the amplitude reached 2^-(1/2) of the local maximum was measured.]]
-
|[[Image:ETHZ_Basel_models_evaluator_ip_mello_frl.png|thumb|400px|'''Evaluation of activation (FRL time constant as predicted by the model from Mello & Tu (2003): 0.3205s''' Time elapsed upon activation of the system until the amplitude reached 2^-(1/2) of the local maximum was measured.]]
+
|[[Image:ETHZ_Basel_models_evaluator_ip_mello_frl.png|thumb|400px|'''Evaluation of activation (FRL time constant as predicted by the model from Mello & Tu (2003): 0.2985s''' Time elapsed upon activation of the system until the amplitude reached 2^-(1/2) of the local maximum was measured.]]
|}
|}
-
The similarity of the predictions of both models suggests a time constant around 0.33s. Both plots show knicks (especially the Mello & Tu (2003) prediction) at the time, when the light pulses were turned off. The reason for this is the immediate cut off of species flow in both direction and following re-equilibration of the system. This happens for both red and far-red light pulse induction.
+
The similarity of the predictions of both models suggests a time constant around 0.35s. Both plots show kinks (especially the Mello & Tu (2003) prediction) at the time, when the light pulses were turned off. The reason for this is the immediate cut off of species flow in both direction and following re-equilibration of the system. This happens for both red and far-red light pulse induction.
{| border="0" align="center"
{| border="0" align="center"
|- valign="top"
|- valign="top"
-
|[[Image:ETHZ_Basel_models_evaluator_ip_spiro_rl.png|thumb|400px|'''Evaluation of inactivation (RL) time constant as predicted by the model from Spiro et al. (1997): 0.2200''' Time elapsed upon activation of the system until the amplitude reached 2^-(1/2) of the local maximum was measured.]]
+
|[[Image:ETHZ_Basel_models_evaluator_ip_spiro_rl.png|thumb|400px|'''Evaluation of inactivation (RL) time constant as predicted by the model from Spiro et al. (1997): 0.1524s''' Time elapsed upon activation of the system until the amplitude reached 2^-(1/2) of the local maximum was measured.]]
-
|[[Image:ETHZ_Basel_models_evaluator_ip_mello_rl.png|thumb|400px|'''Evaluation of inactivation (RL time constant as predicted by the model from Mello & Tu (2003): 0.1380''' Time elapsed upon activation of the system until the amplitude reached 2^-(1/2) of the local maximum was measured.]]
+
|[[Image:ETHZ_Basel_models_evaluator_ip_mello_rl.png|thumb|400px|'''Evaluation of inactivation (RL time constant as predicted by the model from Mello & Tu (2003): 0.1305s''' Time elapsed upon activation of the system until the amplitude reached 2^-(1/2) of the local maximum was measured.]]
|}
|}
-
The red light is faster than the far-red light reaction. Nevertheless, the time scale is very similar and the mean of 0.18 seems reasonable.
+
The red light is faster than the far-red light reaction. Nevertheless, the time scale is very similar and the mean of 0.14 seems reasonable.
 +
 
 +
== Download ==
 +
The evaluation script for experimental design requires the [[Team:ETHZ_Basel/Achievements/Matlab_Toolbox|Matlab Toolbox]]  to be installed and can be downloaded from the [[Team:ETHZ_Basel/Achievements/Systems_Design|Achievements/Systems Design]] section.

Latest revision as of 17:52, 27 October 2010

Experimental Design

Since the design of E.lemming implies the existence of many possible biological combinations of the fundamental parts, the first steps in the modeling of our system were directed towards reducing the actual number of combinations implemented in the wetlab. Furthermore, by using the combined molecular models for in silico evaluation of the best possible devices, we derived theoretical results on choosing the biological parts which maximize the chance of a functional final ensemble.

Wet laboratory evaluation results have showed that molecular modeling and experimental biology can interwork to gain new insight on both aspects of our project.

Models used for experimental design. Coupled individual models for the simulation of the whole process and their interfaces.

Insights for wet laboratory

Goals of the evaluation

To determine the best network structure for the biological implementation of the E. lemming and the best experimental conditions, the combined molecular model (PhyB/PIF3 light switch - Chemotaxis) has been used to answer specific questions:

  1. Which protein of the chemotaxis pathway (CheR, CheB, CheY or CheZ) should be fused...
  2. ...to which light-sensitive protein (PhyB or PIF3)?
  3. Which aspartate concentration should be used in the experiments?
  4. Which concentrations of the two synthetic fusion proteins (Che-LSP1, Tet-LSP2) and the DNA binding sites (tetO) result in a high sensitivity of the network on red and far-red light?

Answers to this questions were able to decrease the experimental expenses significantly.

Evaluator

Determination of relative amplitude. Relative amplitude of free CheYp was maximized to determine optimal parts.

Evaluation of the results is a general optimization problem. The output variable in the case of E. lemming is the concentration of free and phosphorylated CheYp, because it is directly linked to the movement bias. To achieve a reasonable evaluation variable, the relative amplitude of free and phosphorylated CheYp was chosen to be maximized.

Because of the limited amount of different combination possibilities, it was decided to evaluate all combinations. Parameter space of variable experimental implementation possibilities is shown in Table 1.

Table 1: Experimental parameters
Che LSP1 LSP2 [Asp] [AP] [anchor] model
CheR PhyB PIF3 0 μM 0 μM 0 μM Spiro
CheB PIF3 PhyB 10-6 μM 25 μM 25 μM Mello
CheY 10-3 μM 50 μM 50 μM
CheZ 75 μM 75 μM
100 μM 100 μM

The questions were answered in a hierarchical order, e.g. try Che1 before Che2.

Results

1. Which protein of the chemotaxis pathway (CheR, CheB, CheY or CheZ) should be attacked?

Che = CheR.

Amplitude: 0.35
(CheR, LSP1:PIF3, Asp:Medium, Model:Spiro)
Che = CheB. Inverse relation of activation / inactivation of CheB limit possible applications in E. lemming.

Amplitude: 0.35
(CheB, LSP1:PIF3, Asp:Medium, Model:Spiro)
Che = CheY.

Amplitude: 0.9
(CheY, LSP1:PIF3, Asp:Medium, Model:Spiro)
Che = CheZ. CheZ showed the most increase in terms of relative amplitude.

Amplitude: 1.6
(CheZ, LSP1:PIF3, Asp:Medium, Model:Spiro)

Although CheZ showed the highest relative amplitude, it was not chosen to be the first candidate, since it is located at the cell membrane and therefore, CheY would still be able to dephosphorylate CheZ, because CheY is not strictly spatially located. For this reason, CheY was chosen to be the first target. CheR and CheB would have been possible backup targets in this order.

In addition, CheB and CheZ showed inverse activation / inactivation behavior than CheR and CheY upon light pulse induction. This means, that the activity would increase and therefore, this would have to be corrected by increasing the concentration of the proteins far above wild-type level to achieve an inverse effect of activation.

2. To which light-sensitive protein (PhyB, PIF3) should the Che protein be linked?

LSP1 = PhyB. PhyB coupling is very robust and results in maximal plateau surface.

Amplitude: 0.5
(CheY, LSP1:PhyB, Asp:Medium, Model:Spiro)
LSP1 = PIF3. Coupling to PIF3 results in a much more sensitive area of [AP] vs. [anchor].

Amplitude: 0.9
(CheY, LSP1:PIF3, Asp:Medium, Model:Spiro)

PhyB is according to the evaluation the more robust choice compared to PIF3, but it's speculated that PhyB might have the tendency to sequester, since PhyA does under certain conditions. This would result in a lowered active concentration of the species, because the proteins stick together. In addition, PIF3 is much smaller than PhyB and it could therefore have a lower negative influence on the Che protein. Relative amplitude of PIF3 is higher, although at lower levels, it's very similar. For this reasons, PIF3 was the first choice as light-sensitive protein linked to the Che protein.

3. What is a good aspartate concentration?

[Asp] = 0 μM. No aspartate in the systems results in very low relative amplitude.

Amplitude: 0.03
(CheR, LSP1:PIF3, Asp:Zero, Model:Spiro)
[Asp] = 10^-6 μM. Medium aspartate levels show reasonable amplitude levels of amplitude.

Amplitude: 0.35
(CheR, LSP1:PIF3, Asp:Medium, Model:Spiro)
[Asp] = 10^-3 μM. High aspartate levels reach saturation of receptors and cannot increase the relative amplitude much further.

Amplitude: 0.5
(CheR, LSP1:PIF3, Asp:High, Model:Spiro)

Aspartate concentration should be chosen to reach saturation or near-saturation of aspartate receptor. In this example, medium aspartate levels were able to reach near-saturation of the receptors and therefore, this parameter value was chosen.

4. In what quantity should the anchor and binding light-sensitive protein be chosen?

Spiro prediction with first choice parameters.

Amplitude: 0.9
(CheY, LSP1:PIF3, Asp:Medium, Model:Spiro)
Mello prediction with first choice parameters.

Amplitude: 0.9
(CheY, LSP1:PIF3, Asp:Medium, Model:Mello)

According to the the evaluation of all 48 parameter combination possibilities, it was decided to take a concentration for anchor plasmid of [anchor] = 50 μM and a concentration for the anchor binding protein [AP] = 40 - 50 μM, which is slightly below to ensure having enough binding positions.

Comparison of the chemotaxis models

Both chemotaxis receptor models implemented in the combined model predicted very similar results. An obvious difference is the missing consideration of CheZ in the Mello model.

Insights for information processing

Goal of the evaluation

Main goal of the evaluation for information processing was to determine time constants of far-red light and red light induction and resulting change in E. lemming by using the combined molecular model (Light switch - Chemotaxis). These results are the experimentally optimal light pulse times for the LEDs attached to the controller, as predicted by our molecular model.

Evaluator

Again, the output variable was the concentration of free and phosphorylated CheYp. To determine the time constants, the time elapsed upon activation of the system until the amplitude reached 2-1/2 of the local maximum was measured. Results were determined for our target configuration of E. lemming (CheY, LSP1:PIF3, Asp:Medium).

Results

Evaluation of activation (FRL) time constant as predicted by the model from Spiro et al. (1997): 0.4092s Time elapsed upon activation of the system until the amplitude reached 2^-(1/2) of the local maximum was measured.
Evaluation of activation (FRL time constant as predicted by the model from Mello & Tu (2003): 0.2985s Time elapsed upon activation of the system until the amplitude reached 2^-(1/2) of the local maximum was measured.

The similarity of the predictions of both models suggests a time constant around 0.35s. Both plots show kinks (especially the Mello & Tu (2003) prediction) at the time, when the light pulses were turned off. The reason for this is the immediate cut off of species flow in both direction and following re-equilibration of the system. This happens for both red and far-red light pulse induction.

Evaluation of inactivation (RL) time constant as predicted by the model from Spiro et al. (1997): 0.1524s Time elapsed upon activation of the system until the amplitude reached 2^-(1/2) of the local maximum was measured.
Evaluation of inactivation (RL time constant as predicted by the model from Mello & Tu (2003): 0.1305s Time elapsed upon activation of the system until the amplitude reached 2^-(1/2) of the local maximum was measured.

The red light is faster than the far-red light reaction. Nevertheless, the time scale is very similar and the mean of 0.14 seems reasonable.

Download

The evaluation script for experimental design requires the Matlab Toolbox to be installed and can be downloaded from the Achievements/Systems Design section.