Production of tumor antigen attached OMV through E.coli Top10 strain
The design of the outer membrane vesicle (OMV) protein outlined in figure 1 contains three major structural domains: vesicle surface protein C terminus domain ClyA, cancer cell antigen Adpgk domain, and mouse IgG Fc fragment for the purpose of immune-enhancing and presenting antigen.
In order to verify the expression of our OMV protein, a plasmid based on pET28b vector was constructed and used to transform to E.coli BL21 DE3. Plasmid was extracted from successful transformants and sequenced to check for fidelity. Different concentrations of IPTG were tested to determine the concentration required for induction and SDS-PAGE was used to detect the presence of OMV protein in supernatant of culture media, bacteria lysate and pellet. (experiment detail on: supplementary material: protocol 7,8,12).
Repeated experiments were unable to detect the presence of the OMV through SDS-PAGE as shown in figure 3. Figure 3c shows weak band around 95kDa was observed in pellet of the bacterial lysate. This initially promising signal at the correct molecule mass. However, this signal was not reproducible and so insufficient.
Discussion with Prof Huang led to the development of an alternative detection technique with higher sensitivity. The plasmid was redesigned to include a HA (hemagglutinin) tag for Western Blotting and, an enhanced GPR reporter gene. The design of the redesigned plasmid is shown in figure 4.
E.coli BL21 DE3 were transformed with the redesigned pET28b vector and transformants were induced with 0mM - 2mM IPTG and screened for GFP expression. (experiment detail on: supplementary material: protocol 12).
After induction at 37℃ for 4h a significant fluorescence signal was detected and is show in figure 5. The optimum IPTG concentration for expression was determined to be in the range of 0.4mM to 0.6mM. Figure 6 displays images of E.coli transformants expression enhance GFP.
The signal of GFP florescent protein is indicative of OMV protein expression. It was decided to perform a western blot to confirm the weak band noted in figure 3c.
Figure 7 shows the western blot image confirming the band on Fig3c is from the OMV. It was determined that the whole system was working. (experiment detail on: supplementary material: protocol 9).
Control of OMV release through UV light
To overcome the immune tolerance caused by long-term antigenic stimulation, OMV released was controlled using the SulAp promoter (BBa_K518010), an optogenetic regulatory element that was previously used by our team in 2020.
After transforming the vector into E.coli Top10 the bacteria were exposed to UV radiation (15mW/CM2) for 1min and tested for change of the florescence. (experiment detail on: supplementary material: protocol 13)
The florescence of bacteria culture increases significantly after exposure to UV. It was decided to perform further testing through western blot. (experiment detail on: supplementary material: protocol 9)
Redesign of part sulAp and reduce leaking expression
The high leaking expression of SulAp make us failed to reach the goal of constructing a UV induced bacteria lysis system in 2020(https://2020.igem.org/Team:SZ-SHD/Proof_Of_Concept). By redesigning the part sulAp, we significant reduce the leaking expression without a large affect on the promoting strength.
Detail information on: https://2022.igem.wiki/sz-shd/improve
The Deep Nero Network
The Nero Network was made and trained as a classification model, inputting the image (marked as X) and outputting the possibility distribution of the classes. More information regarding the code and the architecture is outlined on the modelling page and the Software tool page. The results are outlined below. The textbox below shows the classes arranged in this order
upper-gi-tract\pathological-findings\esophagitis-b-d 0
lower-gi-tract\therapeutic-interventions\dyed-lifted-polyps 1
upper-gi-tract\anatomical-landmarks\z-line 2
lower-gi-tract\pathological-findings\ulcerative-colitis-grade-2-3 3
lower-gi-tract\pathological-findings\ulcerative-colitis-grade-1-2 4
lower-gi-tract\pathological-findings\ulcerative-colitis-grade-0-1 5
upper-gi-tract\pathological-findings\barretts 6
upper-gi-tract\anatomical-landmarks\pylorus 7
upper-gi-tract\anatomical-landmarks\retroflex-stomach 8
upper-gi-tract\pathological-findings\barretts-short-segment 9
lower-gi-tract\pathological-findings\ulcerative-colitis-grade-1 10
lower-gi-tract\pathological-findings\ulcerative-colitis-grade-2 11
lower-gi-tract\anatomical-landmarks\ileum 12
lower-gi-tract\anatomical-landmarks\retroflex-rectum 13
lower-gi-tract\quality-of-mucosal-views\impacted-stool 14
lower-gi-tract\pathological-findings\polyps 15
lower-gi-tract\anatomical-landmarks\cecum 16
upper-gi-tract\pathological-findings\esophagitis-a 17
lower-gi-tract\pathological-findings\hemorrhoids 18
lower-gi-tract\therapeutic-interventions\dyed-resection-margins 19
lower-gi-tract\pathological-findings\ulcerative-colitis-grade-3 20
lower-gi-tract\quality-of-mucosal-views\bbps-0-1 21
lower-gi-tract\quality-of-mucosal-views\bbps-2-3 22
e.g. class 18 is the "lower-gi-tract\pathological-findings\hemorrhoids" class
Heatmap
Figure 1 shows the lightness of the pixel with index of [Row = r, Column = c] means the number of the sample been classified as a c-th class and it's acctually the r-th class According to the diagram, class 23 has a lot of wrong data, because class 23 is the class for unpredictables
Figure 2 shows the same thing but less sophisticated. In this image, the lightness of the pixel with index of [Row = r, Column = c] therfore
Where X is a sample from the sample space and 
is the prediction from the network
To compare with other networks or the same network with different hyperparameters, or just curiosity, here are the metrics
| accuracy_score | accuracy_score | recall_score | f1_score | |
|---|---|---|---|---|
| The model | 0.7297880322641156 | 0.7297880322641156 | 0.976408583260133 | 0.8352745424292845 |
The results for this model are really decent and as a medical-purposed model. Because the higher the recall is, the safer it is, and therefore the 0.97 recall (which is very high and very good) is so satisfying.
To deploy the model in production, we have to add the corresponding timing information( which we wont have the data until we do the animal test, because endoscopys don't move the way capsules move) to it with RNNs or transformers. But it's clear enough to say the data will just be better because better data is fed in.