1.Building cell factories at the level of creating new biological systems
The use of rapidly developing synthetic biology is an effective strategy for building cell factories. The core of synthetic biology is to design and construct new biological systems or new life forms with physiological functions to achieve sustainable green biomanufacturing, and three levels of cell factory construction have been proposed: standardized components/dynamic regulation and the construction of new life systems. Meanwhile, the creation of new life systems is precisely as the ultimate goal of synthetic biology.
//img_wrap
Fig 1. Gene editing in Escherichia coli chassis cells using the crisper technique（Photo credit: Tianyuan Su. A CRISPR-Cas9 Assisted Non-Homologous End-Joining Strategy for One-step Engineering of Bacterial Genome.Sci Rep.2016）
//img_wrap
In current synthetic biology research, minimalist strains are often used as the starting research object because they can bring: lower genome complexity and other excellent characteristics. However, no researcher has proposed how to combine the advantages of prokaryotic streamlining with those of eukaryotic polyploidization.
In this project, our team, SDU-CHINA, followed the “To build to understand” principle of synthetic biology and asked the following questions: Is it possible to create a new polyploid lean prokaryotic cell? What is the phenotype of this cell? Will it have the advantages of both? What can it tell us about evolution?
//img_wrap
Fig 2. Designing and constructing polyploid Escherichia coli by fine-tuning FtsZ expression levels
//img_wrap
We introduced a polyploid based on a prokaryotic lean strain by regulating the expression intensity of the cell division gene ftsz, and finally succeeded in constructing a polyploid minimalist Escherichia coli (PMEC).
This may provide some insight for subsequent teams participating in igem, and attempts to build new life systems based on the introduction of standardized components and dynamic systems could broaden the level of building cell factories in synthetic biology.
2. Improving COBRA metabolic network models
2.1 A multi-objective optimization method for simulating bacterial metabolic behavior using a combination of COBRA metabolic network models.
In the process of solving the distribution of metabolic fluxes, defining only a single objective function can lead to various issues, such as an inability to simulate the reasonable allocation of metabolic fluxes by bacteria. For example, if we define the production of PHB as the sole objective function, the model may maximize the reaction rate for PHB production at the expense of the biomass reaction rate, which could be assigned a value of 0 mmol/gCDW/h. This is clearly unreasonable (see Fig. 3a in the modeling section).
Therefore, we propose a multi-objective optimization approach to simulate the metabolic flux behavior of bacteria. The modified objective functions are as follows:
\[F\left(x\right)=\omega_1v_1+\omega_2v_2\ +\ ...\ +\ \omega_nv_n\] \[Maximize\ Fx= ω^T v\]where v1,v2,v3… are the reactions we want to define. For instance, we can simultaneously define the PHB production reaction and the biomass reaction. This simulation approach reflects the balance between bacterial growth and production (see modeling Part 1.2 for details). The simulation results are shown below.
//img_wrap
Fig 3. Results of multi-objective function optimization, where the model considers both PHB synthesis and individual growth.
//img_wrap
IGEM follow-on teams can download EASYCobra from our gitlab, its a metabolic network analysis python package that integrates a lot of functionality, including the above features
2.2 A method for integrating transcriptome data with the Cobra metabolic network model.
The method for flux balance analysis based on the metabolic network model established by Cobra is as follows, where S represents the stoichiometric matrix of the metabolites, and V is the flux matrix for each reaction.
\[S\ast V=0\]The differential expression ratios for each gene were determined through transcriptome analysis, and we transformed these expression ratios into a weight matrix W. Consequently, the modified flux balance analysis is presented as follows:
\[S\ast V ∘W=0\]The entire process is illustrated as follows:
//img_wrap
Fig 4. Integration of transcriptomic data with the GSSM of DGF-298
//img_wrap
IGEM follow-on teams can download EASYCobra from our gitlab, its a metabolic network analysis python package that integrates a lot of functionality, including the above features.
2.3 A user-friendly metabolic network analysis python package based on Cobra
EasyCOBRA is a Python package we developed that provides numerous methods for analyzing and modifying metabolic networks, while being more user- friendly. It consists of two components: EasyCobraModifier and EasyCobraAnalyzer. EasyCobraModifier provides a convenient way to modify metabolic networks and detect reaction equilibrium. EasyCobraAnalyzer provides analytical methods such as multi-objective optimization and visualization of metabolic networks.
Future igem teams can download EasyCOBRA in our gitlab.
3. A deep learning method combine multi-sensor for predicting PHB yield in fermentation tanks.
We trained a deep learning model based on LSTM that predicts PHB yield in real-time by utilizing easily measurable data during the fermentation process, such as PHB, temperature, sugar consumption, and OD used in this training (see modeling Part 2.1 for details). The schematic diagram is shown in Figure 5.
//img_wrap
Fig 5. Real-time Prediction of Products by Combining Multi-Sensors with Deep Learning Models.2.2: Model Training and Evaluating
//img_wrap
The loss of the model during the training phase is shown in Fig. 6, where the loss reached 0.000174 after 100 epochs. In the validation set of the third fermentation, the accuracy achieved 0.9871, and the mean squared error was 0.19 (Fig. 6).
//img_wrap
Fig6. Relationship between Loss and Epochs.
Fig7. prediction result
//img_wrap
The evaluation metrics of the model.//caption
| Metric | Value |
|---|---|
| Mean Square Error (MSE) | 0.1914 |
| Root Mean Square Error (RMSE) | 0.4375 |
| Mean Absolute Error (MAE) | 0.3833 |
| R2 Score | 0.9871 |
At the same time, we have designed a multi-sensor hardware（Fig. 14） which can measure CO2, temperature, air pressure and humidity by the end of the project. We will continue to improve the hardware.
//img_wrap
Fig 7 Hardware physical drawing (left), blueprint (right)
//img_wrap
We demonstrate that the combination of multi-sensor designed in conjunction with the model, with sufficient data point inputs, enables the prediction of the whole fermentation process for a variety of chassis cells for different high value-added products.
How to organize a Focus Group
//img_wrap
Fig 8 Focus Group process
//img_wrap
First, clearly define your research objectives and questions. Identify the information you aim to gather through the Focus Group and formulate specific questions to guide the discussion and address your research goals. Second, select suitable participants, ensuring they represent your target group by considering factors such as age, gender, and background. Ideally, strive for diversity within the group. Next, develop a guiding outline that covers the main topics of discussion. Create a relaxed atmosphere that encourages participants to speak freely. Finally, take notes and summarize the discussion for in-depth analysis.
Some tools to help iGEM teams build their wiki
During the wiki coding procedure, we have tried to make every parts reusable and migratable. After all work finished, we want to share it with future teams and help them build their team wiki. So this page is here to give instructions about the tools as well as provide some of my experiences and opinions.