Line 99: | Line 99: | ||
<p>A pattern recognition example in computer simulation is shown here. The DNA-based neuron is trained to recognize a 'T' in a 3\(\times\)3 grid. The grey scales of the nine grids are provided and represented by the concentrations of nine species. Also, there are nine weights corresponding to the nine inputs. During training, if the image should be recognized as 'T', we provide a `desired answer' species which has high concentration. Otherwise the concentration of the species is set to 0.</p> | <p>A pattern recognition example in computer simulation is shown here. The DNA-based neuron is trained to recognize a 'T' in a 3\(\times\)3 grid. The grey scales of the nine grids are provided and represented by the concentrations of nine species. Also, there are nine weights corresponding to the nine inputs. During training, if the image should be recognized as 'T', we provide a `desired answer' species which has high concentration. Otherwise the concentration of the species is set to 0.</p> | ||
− | <p>The figure below shows the training process of the neuron. We present the input image during each training (the upper grids) and the corresponding weights within the neuron (the lower grids). The results are shown at the bottom of the figure. Only after 10 times of training, the neuron can successfully recognize the target 'T'. More detaied information can be found in the following | + | <p>The figure below shows the training process of the neuron. We present the input image during each training (the upper grids) and the corresponding weights within the neuron (the lower grids). The results are shown at the bottom of the figure. Only after 10 times of training, the neuron can successfully recognize the target 'T'. More detaied information can be found in the following tables:</p> |
<table> | <table> | ||
<tr> | <tr> | ||
Line 112: | Line 112: | ||
<td>10</td> | <td>10</td> | ||
<td>0.8</td> | <td>0.8</td> | ||
+ | </tr> | ||
+ | </table> | ||
+ | <table> | ||
+ | <tr> | ||
+ | <td></td> | ||
+ | <td>w\(_1\)</td> | ||
+ | <td>w\(_1\)</td> | ||
+ | <td>w\(_1\)</td> | ||
+ | <td>w\(_1\)</td> | ||
+ | <td>w\(_1\)</td> | ||
+ | <td>w\(_1\)</td> | ||
</tr> | </tr> | ||
</table> | </table> |
Revision as of 11:32, 19 October 2019
![](https://static.igem.org/mediawiki/2019/b/be/T--SEU--tm-wy1.jpg)
Demonstrate
Overview
This year, the contribution of our project mainly lies in a theory and its validation in molecular computing. Our project consists of three parts: a general computation theory in molecular computing, a software tool and the corresponding experimental validation of such parts.
This page presents the results of dry experiments (which prove our theory using in silico simulation) and wet experiments (in which materials are generated by our software tool).
Dry Experiments
This part presents results of our Model.
1.The simulation results of each calculation operation.
Addition:
The figure below shows the numerical simulation result of a set of reactions:\(A_1 \xrightarrow{k} O,\quad A_2 \xrightarrow{k_2} O, \quad A_3 \xrightarrow{k_3} O\) which perform addition calculation. The initial concentrations (input values) are 1, 2 and 3, respectively (dashed lines in the figure). The output result is the sum of such values (solid red line in the figure).
![](https://static.igem.org/mediawiki/2019/8/84/T--SEU--additionSim.png)
Subtraction:
The figure below shows the reaction \(A+B \xrightarrow{k} \phi\) which is a subtractor. Experiment settings in this figure is: \([A](0)=2, [B](0)=3\).
![](https://static.igem.org/mediawiki/2019/9/93/T--SEU--subtractionSim.png)
Multiplication:
The numerical results of reactions \(\alpha \xrightarrow{k_1} \phi, A+B+\alpha \xrightarrow{k_2} A+B+\alpha+C\) are shown in the figure below. Initial concentrations are \([A](0)=4, [B](0)=3\). The result shows that the final concentration of \(C\) reaches \(4 \times 3=12\).
![](https://static.igem.org/mediawiki/2019/b/bb/T--SEU--multiplicationSim.png)
2.We use such a model to construct a chemical neuron.
A pattern recognition example in computer simulation is shown here. The DNA-based neuron is trained to recognize a 'T' in a 3\(\times\)3 grid. The grey scales of the nine grids are provided and represented by the concentrations of nine species. Also, there are nine weights corresponding to the nine inputs. During training, if the image should be recognized as 'T', we provide a `desired answer' species which has high concentration. Otherwise the concentration of the species is set to 0.
The figure below shows the training process of the neuron. We present the input image during each training (the upper grids) and the corresponding weights within the neuron (the lower grids). The results are shown at the bottom of the figure. Only after 10 times of training, the neuron can successfully recognize the target 'T'. More detaied information can be found in the following tables:
Number of weights | Initial value of weights | Training times | Recognition threshold |
9 | 0.3 | 10 | 0.8 |
w\(_1\) | w\(_1\) | w\(_1\) | w\(_1\) | w\(_1\) | w\(_1\) |
![](https://static.igem.org/mediawiki/2019/5/5c/T--SEU--pattern.png)
Wet Experiment
1. Kinetics experiments of addition
The PAGE imaging result of reactant complexes and reactants of addition reaction.
![](https://static.igem.org/mediawiki/2019/c/c1/T--SEU--additionPAGE.png)
The fluorescence intensity of the one-step reaction system:
The fluorescence intensity of the first step of the two-step reaction system:
The fluorescence intensity of the second step of the two-step reaction system:
In order to demonstrate the reaction really happened, we used 12% native PAGE to visualize the DNA fragments.
![](https://static.igem.org/mediawiki/2019/a/a9/T--SEU--additionPAGEDemonstrate.png)
And we could observe the formation of Oi (the target product of the first step of two-step group), two kinds of wastes, and the mixture of b and c (the target product of the whole experiment). We could see from the stripes of the first step that with the formation of Oi, which was about the same height of 1Gia (in this reaction Oi equals to 1Gia), another light stripe occurred below the 50 bp marker of dsDNA marker, which should be the waste of the first step of addition. The products, ssDNA b and ssDNA c share the same length of 40 bp, but were not very legible because signals of ssDNAs stained by GelRed were naturally much smaller than dsDNAs . However, we could easily recognize the stripes of two wastes of this reaction, which were all double-stranded DNAs.
As analyzed above, there can be something wrong with the experiments. We tried the qRT-PCR experiments again by using another two machines (Applied Biosystems QuantStudio 3, Applied Biosystems StepOnePlus Real Time PCR System). The reaction systems were the same as above experiments, but with 1.4 uL ROX reference dye added in every reaction system conducted by Applied Biosystems QuantStudio 3 and Applied Biosystems StepOnePlus Real Time PCR System to reduce error. After analysis and discussion, we referred to the latter one as the correct machine which could provide us trustworthy data.
However, a slight ascendance at the beginning of PCR still existed. We hypothesized that it attributed to the combination of fluorescent 1Gia and the quencher 1Tic, that is, the unreacted single stranded DNAs in reactant complexes when 1Gi and 1Ti were mixed before PCR. To reduce the interference of unreacted single stranded DNAs in reactant complexes, we conducted purification to the reactant complexes (1Gi and 1Ti in addition reaction) after above experiments. The results are as follows:
2. Kinetics experiments of subtraction
The PAGE imaging result of reactant complexes and reactants of subtraction reaction.
![](https://static.igem.org/mediawiki/2019/a/a6/T--SEU--subtractionPAGE.png)
References
[1] CRNsimulator.