Team:NCTU Formosa/Proof Of Concept
Loading...
Introduction
Concept
DenTeeth can produce antimicrobial peptides, LL-37 when the concentration of bacteria in the mouth is higher. After the growth of bacteria is inhibited, STATH and BMP2 will express, maintaining a high calcium level in saliva, and repairing soft tissues in the oral cavity. Therefore, oral problems, especially periodontal disease can be successfully prevented.
How do we prove it?
We proved our concept with a meticulous process which can be roughly divided into three parts: Model, Lab Work, and Device design. Combining modeling results and predictions with our lab work, we enable to make DenTeeth work as we imagined. We could further prove that DenTeeth can be implemented in the real world for daily usages.
Bacteria Growth Simulation with DenTeeth
See the Prediction Model page for more information!
Considered the previous growth model plus the killing formula of LL37. We constructed model to simulate the growth curve of E. coli and P.gingivalis with DenTeeth.
As we can see above, the concentration of P. gingivalis and E. coli are reduced, and finally, they will achieve dynamic balance, which is the result we needed. After we know our assumption is possible, we need more information about our protein production.
LL-37 tetR mRFP and BMP2 STATH GFP Production Simulation
See the Prediction Model page for more information!
Because E. coli itself would also be affected by LL37, in order to test whether this will further affect the concentration of the target product, we then used the analysis above to predict the concentration of these products over time.
The total amount of AHL was composed of the initial AHL from the quorum sensing model. The AHL-LuxR complex would activate the Plux promoter , which could lead to the production of LL37, tetR and mRFP.
When the concentration of bacteria was low, DenTeeth would start to produce BMP2, STATH and GFP. Thus, we wanted to predict the production of these proteins. Considering the Quorum Sensing Model, we constructed the simulation of BMP2 and STATH.
In order to observe the switching between sterilization and restoration of DenTeeth, we added RFP after the sterilization sequence and GFP after the restoration sequence. Next, we simulated the relative fluorescence intensity of RFP and GFP to know the actual operation of DenTeeth. The result is shown in the figure below. :
These models help us to calculate and simulate the experiment result. Next, we will use the data from experiments to validate the model.
Model Validation
See the Prediction Model page for more information!After doing Prediction Model Validation Test, we use data from it to validate our model. You can find more details in Prediction Model page, and we get the validated prediction as below figure.
And we also predicted the new growth curve of E. coli and P.gingivalis with DenTeeth. As you can see in the picture (Fig.13), compared to the growth of P.gingivalis without DenTeeth (Fig.1), the final O.D. value of P.gingivalis decreased from 0.7 to 0.08, which showed that our DenTeeth could effectively kill 88% of the pathogenic bacteria in dogs' oral cavities.
The validation result is a fascinating result for us. From these, we can make sure that DenTeeth can kill up to 88% of the pathogenic bacteria in dogs' oral cavities. But besides proving the feasibility, we still need another experiment to know whether we succeed in producing our proteins.
Protein Functional Test
To know whether its inhibition ability can have a function, we did the following experiment. You can find more details in the Result page.
LL-37 Functional Test
We use the inhibition zone experiment to confirm that DenTeeth can inhibit other bacteria. Furthermore, we can also compare the difference of inhibition zone diameter to know the strength of inhibition intensity.
These results successfully demonstrate the function of LL-37 we produced. After this, we want to know the one of STATH.
STATH Functional Test
STATH can prevent the precipitation of calcium phosphate in saliva and maintain a high calcium level in saliva. Therefore, STATH is available for promoting the remineralization of tooth enamel and preventing calculus formation. So we do the related experiment to prove our assumption, and you can see the results as below:
In conclusion, LL-37 and STATH are successfully produced and have their functions to reach our target, so we want to introduce you to the idea of how they are utilized in our daily life.
Efficiency Optimization Model
Our Efficiency optimization model combines reinforcement learning and takes our prediction model as the environment to virtualize the interaction among our designed dental bone, P. gingivalis in dog mouth, and the reinforcement learning model.
To investigate if the RL model shows better efficiency, we do a comparison with fixed time feeding. Figure 12.A shows the reward of the RL model and fixed time feeding with respect to time. From figure 12.A, we found that the model is capable of increasing the reward as the episode goes. Figure 12.B is the cumulative reward with respect to time of RL model and fixed time feeding. We are able to observe that as the episode goes, the optimization efficiency of the RL model is gradually greater.
DenTeeth-bone
See the Implementation page for more information!
Design
We designed DenTeeth-bone into two main structures, the comb-like shape one and the elliptical cylinder one. The comb-like shape is to increases the contacting area, which can improve the efficiency of using DenTeeth-bone. The elliptical cylinder structure is made as a grip for the owner to hold DenTeeth-bone tightly when the dog is chewing on the comb-like structure.
We recommend the owner to hold on to the elliptical cylinder structure when feeding DenTeeth-bone, this will help the dogs' teeth rub more with the DenTeeth-bone and increase the cleaning effect. After the dog ate up the comb-like structure, the owner can unloose the grip and let the dog consume the elliptical cylinder part happily.
Precise feeding of DenTeeth-bone
Since DenTeeth-bone is a consumable product and eating too many DenTeeth-bones might have some negative impacts on the dog's health, we have to calculate the best frequency of using DenTeeth-bone.
However, because the oral environment in different dogs fit different conditions, it will be difficult to figure out the best frequency of using DenTeeth-bones for different dogs.
To achieve the goal, we constructed the reinforcement learning model. The goal of the model is to use the least DenTeeth-bones to achieve the best effect of sterilization.
The model will calculate the best frequency of feeding and tell the user when should they feed DenTeeth-bone. By this model, the DenTeeth-bones we produce can be successfully applied to all types of dogs.
Reference
- Schink, S. J., et al. (2019). "Death rate of E. coli during starvation is set by maintenance cost and biomass recycling." 9(1): 64-73. e63.