Our model is divided into three parts:
bioinformatics analysis ,
computational chemistry
mathematical model.

Screening of Targeting Fragments
To achieve bacteria-bacteria targeting, we devised two targeting approaches involving calcium-binding proteins and pilus self-assembly. Through protein-protein docking and molecular dynamics simulations, we ultimately opted for pilus self-assembly as the preferred targeting method.

To achieve adhesion to CRC, we separately devised strategies involving pilus adhesion, antibody binding, and HlpA adhesion. Considering wet lab results, cost, and experimental timeframes, we ultimately selected HlpA as the preferred approach for targeting CRC.

Rational Protein Design
In order to enable engineered bacteria to target Fn and CRC, we designed two fishing rod proteins and rationalized the modification of membrane proteins. To target and kill Fn and CRC, we designed two cytotoxic peptides, each with its specific focus, and optimized their sequences and spatial structures through protein structure predictions. Based on these designs, we constructed the Dual-Edged Harpoon. Lastly, to enhance the strength of pilus self-assembly, we conducted saturation mutagenesis and rational modification on the pilus monome.

Mathematical Model
Our Dynamics Modeling section ran throughout the project.

We evaluated the killing effect of DEH and the population change of engineered Bifidobacterium (BL), colorectal cancer (CRC) cells, and Fusobacterium nucleatum (Fn). The model considered three spatial situations: the intestinal environment, the tumor surface, and the tumor core.

Applied the Logistic equation for population growth, the Ligand Binding equation for inter-species combination, and the sigmoid Emax model for DEH releasing and killing. The simulation results showed a significant cytotoxic effect within a 10-day timeframe, indicated the potential efficacy of our therapeutic approach.

Mathematical models were solved using R and MATLAB's Simbiology toolbox as a software tool.

Bioinformatics Analysis
Our bioinformatics analysis began with the extraction of key genes characterizing the interaction of Fn with CRC cells from the GEO database. A scoring model was then established based on these genes and applied to a large cohort of CRC patient data from the TCGA database.

RNA-seq analysis was performed to compare gene expression profiles of Fn treated and untreated HT-29 cell lines. DEG and WGCNA analysis helped identify the most correlated modules to Fn. LASSO Cox regression model analysis was used to select the most contributing factors affecting CRC from Fn, resulting in the identification of 11 potential biomarkers. A multivariate Cox model was built and validated using a bootstrap test. The model's predictive accuracy was assessed using a receiver operating characteristic (ROC) curve.

Finally, functional enrichment analysis (FEA) was performed on the 11 genes identified from LASSO, and the tumor immune microenvironment (TIME) was analyzed using the CIBERSORT package to understand the immune perspective.

Our modeling component has provided crucial insights into protein behavior, validated the potential efficacy of our approach, and guided the design of our dual-targeting peptide.