Screening of Fn-targeting fragment
In order to target Fn, we designed four different adhesins and conducted molecular docking for each to assess their binding affinity with mFadA

EC5 was obtained from Protein Data Bank(PDB ID:7STZ). Pep-11 was obtained from Homology Modeling using SWISS-MODEL and the sequence was obtained from literature[1]. The mFadA sequence was obtained from the literature[2] and the structure was predicted by Alphafold2. The B-domain sequence was designed in Engineer Design Circle 2 and the structure was predicted by Alphafold2.

For EC5：Rosetta local_docking，nstruct=10000, repeated 3 times.
For Pep11：Rosie online server[3], RosettaDock-5.0，nstruct=10000, repeated 3 times.
For mFadA：Rosetta local_docking，nstruct=200000, no repeat.
For B-domain：Rosetta local_docking，nstruct=10000，repeated 3 times

The docking results are shown in the following pictures:

Figure 1. Docking results of four Fn-targeting candidates.| a: Diagram of the binding modes of the four peptides. b: Adhesion factors based on self-assembly principle have higher binding energies ( The data were analyzed using student’s T-test; *: P<0.05; **: P<0.01; ***: P<0.001).
Figure 2-1. Molecular dynamics analysis results of EC5.| a. RMSD of EC5 protein complexes; b. Combined Rg and RMSD analysis of EC5 protein complexes. c. RMSF of the EC5 protein complex. d. Analysis of protein slewing in the space of EC5 protein complexes.e~f. Hydrogen bonding between the EC5 protein complex and the solvent and internal hydrogen bonding. g. Covariance matrix of EC5 protein complexes. h. Ramachandran diagram of the EC5 protein complex.( The dramatic fluctuations in the figure are caused by the cyclic boundary of the water box.)
Figure 2-2. Molecular dynamics analysis results of pep11.| a. RMSD of pep11 protein complexes; b. Combined Rg and RMSD analysis of pep11 protein complexes. c. RMSF of the pep11 protein complex. d. Analysis of protein slewing in the space of pep11 protein complexes.e~f. Hydrogen bonding between the pep11 protein complex and the solvent and internal hydrogen bonding. g. Covariance matrix of pep11 protein complexes. h. Ramachandran diagram of the pep11 protein complex ( The dramatic fluctuations in the figure are caused by the cyclic boundary of the water box ).
Figure 2-3. Molecular dynamics analysis results of mFadA.| a. RMSD of mFadA protein complexes; b. Combined Rg and RMSD analysis of mFadA protein complexes. c. RMSF of the mFadA protein complex. d. Analysis of protein slewing in the space of mFadA protein complexes. e. Covariance matrix of mFadA protein complexes. f. Ramachandran diagram of the mFadA protein complex ( The dramatic fluctuations in the figure are caused by the cyclic boundary of the water box ).
Figure 2-4. Molecular dynamics analysis results of B-domain.| a. RMSD of B-domain protein complexes; b. Combined Rg and RMSD analysis of B-domain protein complexes. c. RMSF of the B-domain protein complex. d. Analysis of protein slewing in the space of B-domain protein complexes. e. Covariance matrix of B-domain protein complexes. f. Ramachandran diagram of the B-domain protein complex ( The dramatic fluctuations in the figure are caused by the cyclic boundary of the water box ).

It suggested that adhesion factors based on self-assembly principle have a better ability to bind to bacterial pilus. Therefore, it guided us to consider the display of bacterial pilus monomer or fragment structures on the surface as the basic method for targeting Fn.

Learn from us:

When applying protein engineering for adhesion, binding, or targeting, it is possible to consider the removal of non-essential structural domains to enhance specificity in binding. However, this work should be based on bold assumptions while exercising caution. Because the binding between proteins may involve more than just active sites, the deletion of certain structural domains may lead to irreversible impacts on the correct spatial folding. In future protein structural domain truncation work, computer-assisted protein design can provide comprehensive insights into their structure and function, aiding in better experimental design for wet lab experiments.

Screening of CRC-targeting fragment
In order to target CRC, we designed four different adhesins and conducted molecular docking for each to assess their binding affinity with E-cadherin and HSPG on the surface of CRC.

E-Cadherin and its antibody structure was obtained from Protein Data Bank(PDB ID:7STZ). Based on them, a nanobody and a single chain variable fragment (scFv) were designed by us .

For A-domain：Rosetta local_docking，nstruct=10000，repeated 3 times.
For nanobody：Rosie online server, RosettaDock-5.0，nstruct=1000, repeated 10 times.
For scFv：Rosie online server, RosettaDock-5.0，nstruct=1000, repeated 10 times.
For HlpA docking with HSPG：MOE-2022, ligand-protein, n=1000. repeated 10 times.

Figure 3 Docking results of four CRC-targeting candidates.

The affinities of the A-domain, scFv, and HlpA towards CRC are closely matched. After a comprehensive evaluation, which included considering target expression levels, cost constraints, and experimental timelines, we made the final decision to employ HlpA as the preferred factor for CRC targeting.

Learn from us:

Many times, dry lab and wet lab experiments can be disconnected. The best dry lab results may not necessarily translate to the best outcomes in real-world applications. In computer evaluations, the assessment of something is often based on scores. In contrast, real-world assessments must consider feasibility, cost, robustness, user-friendliness, and various other factors. It must be said that the assistance of dry lab experiments is invaluable in many cases, however, it's crucial not to blindly follow dry lab scores and, instead, make judgments by combining a balanced perspective from various aspects.

Visualization Simulation of Bacterial Interactions.
We conducted simulations in Python based on the BL-Fn adhesion and BL killing Fn designed within the project. We utilized the pygame library to develop two small programs, one for adhesion and one for cytotoxicity, with multiple iterations. Various versions of the logic were developed, all based on the collision of two graphical entities resulting in either adhesion or cytotoxicity. With each iteration, we progressively achieved stable adhesion, consistent cytotoxicity, Brownian motion, periodic boundary conditions, boundary collision rebound, and delayed cytotoxicity, among other features. These simulations allowed us to provide guidance on critical mathematical modeling parameters, such as adhesion rates, adhesion constants, cytotoxicity rate curves, and more, in a two-dimensional context.

However, it's worth noting that many of these functions are not currently compatible, and, to date, we have been unable to merge the adhesion and cytotoxicity programs into a single unified system. Despite recognizing the tremendous potential of this simulation for modeling bacterial interactions, some code issues remain unresolved due to technical limitations.

Therefore, we welcome and encourage interested individuals to engage with us for collaborative learning and further refinement of this simulation system. Our future plans encompass:

1. Combining adhesion and cytotoxicity programs into one.
2. Addressing tunneling issues.
3. Resolving inaccuracies in locating periodic boundaries.
4. Overcoming bottlenecks in simulating interactions between more than two bacterial species.
5. Implementing natural bacterial reproduction and death.
6. Incorporating bacterial chemotaxis and diffusion.
7. Extending bacterial interaction simulations to a three-dimensional context.
8. Discontinuing visualization and focusing on coordinate calculations only, to facilitate faster solutions for mathematical models.