Docking Software Details

Our bioinformatic analysis in drylab was based on the oxalate oxidase from the PDB bank with the ID: 1FI2. The enzyme protein structure was downloaded from the PDB Bank's official website and only the monomer of the hexamer structure was used in the dockings and analysis. This was done for simplicity and reduce the needed computer resources. Ligand files were downloaded from PubChem Explore and prepared using OpenBabel python extensions [3]. The protein preparation was done in ChimeraX using the built-in protein dock prep functionality. The manganese ion was designed a charge of +2.
With the help of the MGL-Tools and ADT Tools the docking made use of many of the provided libraries in these packages [1]. For AutoDock4 we used the Lamarckian algorithm GA(4.2) and our search parameters was Genetic Algorithm function for the docking [2]. The grid point spacing was set to 0.375 angstrom.

Histidine Binding Discovery

One of our initial docking simulations provided insight into the decision of different purification tags. One common purification method for proteins is 6xHis-tags, which are histidine aminoacids in a chain. In one of our docking simulations we discovered that the aminoacid histidine, the one in the purification tag, showed a binding energy of -7.4 kcal/mol around the active site. Hence a free histidine residue from the purification method could interfere with the produced and finished enzyme. To avoid this we opted for a different purification tag. To be specific we instead used a GST-tag (Glutathione S-transferase). This approach minimises the risk of inhibiting our enzyme during analysis of the enzymes activity. The image below shows a histidine molecule binding close to the active site in oxalate oxidase.

AlphaFold Analysis

In parallel, we utilised AlphaFold2 from neurosnap.ai to ensure that the folding of our oxalate oxidase enzyme, along with the added GST and TEV-tags, would occur correctly without any structural interference to the enzyme [4]. We input the amino acid sequence of the enzyme plus the purification tag and verified that the predicted protein structure did not suggest any problems. AlphaFold’s results showed that the enzyme folded naturally, without the GST-tag affecting the active site of the enzyme. Beneath there is a picture of the entire protein that will be translated in the E. coli with our plasmid by its ribosomes.

To further confirm that the enzyme remained functional with the GST-tag attached, we ran a quick docking simulation using the protein bound to the GST-tag and its natural substrate, oxalate. The simulation confirmed that oxalate could still properly access the enzyme's active site and bind effectively. This theoretical validation ensured that our purification method would not interfere with the enzyme’s function, providing confidence in our design for future experiments.

Screening for Inhibitors Within the Digestive Tract

Besides the initial docking simulations, a more extensive and challenging screening was conducted. This aimed to screen all compounds present in the digestive tract, involving over 30 different compounds. The list of compounds were found using perplexity.ai to search for the most common compounds in the digestive tract. The screening process was significantly accelerated using the LightningAutoDock Tools which was developed in drylab. For more information check the page Software in the menu above.

Besides the initial docking simulations, a more extensive and challenging screening was conducted. This aimed to screen all compounds present in the digestive tract, involving over 30 different compounds. The list of compounds were found using perplexity.ai to search for the most common compounds in the digestive tract. The screening process was significantly accelerated using the LightningAutoDock Tools which was developed in drylab. For more information check the page Software in the menu above.

To further investigate the potential of these inhibitors, we utilised our LightningViewer tool to analyse the data and identify compounds with the strongest binding affinities. We found 10 different ligands that exhibited higher binding energies than the remaining. These were then subjected to a targeted docking simulation, as opposed to the blind docking used earlier. The targeted docking screening only checked the active site around the ion.

This targeted docking allowed us to fine-tune the parameters and achieve more accurate results. The refined calculations led to better predictions of the ligand conformations within the active site. As a result, the binding energies are expected to be more aligned with real-world conditions. The increased precision due to the targeted approach compared to blind docking should allow for better estimations.

Below, you can see a diagram of the 10 ligands with the highest initial binding energies, some of which include vitamins, common compounds in the digestive tract, products of the Krebs cycle, and metabolites from lactic acid bacteria. These were the compounds that went through the targeted docking simulation.

Despite the strong binding energies, with some exceeding -10 kcal/mol of free energy in ligand-enzyme interactions, there is limited cause for concern. Many of these larger molecules are unlikely to be problematic, as they would struggle to fit into the active site due to its relatively small opening. This is true for the vitamines that are far too large to fit into the opening of the active site. Additionally, the smaller molecules do not exhibit strong affinities, so they are not likely to pose a significant threat. It’s also worth to mention, that the enzyme oxalate oxidase wouldn’t be secreted extracellularly from the bacteria and therefore these compounds would first have to pass the bacterial cell membrane in order to interact with the protein which minimises the inhibitory risk from these compounds.

As illustrated in the image above, we have mapped the binding of the top 10 compounds, showing how they cluster around the active site. Overall, the screening indicates the presence of potential inhibitors, but they are unlikely to successfully inhibit and bind within the active site.

Optimum pH Prediction

To speed up our work in the wet lab, we used a tool to predict the enzyme's optimal operating pH. This approach helped us significantly reduce the amount of work that would otherwise have been needed. By predicting the optimal pH in advance, we minimised the time spent on trial-and-error experiments, allowing the wet lab team to work more efficiently and complete the experiments faster.
We used a tool called EpHod, which simulates and predicts the enzyme's optimal pH by analysing its sequence and applying deep learning [5]. The tool estimated the optimal pH to be 4.51, a value we then could apply in the wet lab to guide our experiments.