Whole cell models of the Escherichia coli peptidoglycan and the identification of iron-sulfur cluster and zinc binding sites in predicted 3D protein structures

Computer models can allow researchers to examine biological systems in ways that are impossible to do with experiments. This thesis explores modeling biological systems of varying sizes and scales. The first system that I modelled was the bacterial peptidoglycan, which is a net of sugars and peptides that surround the cell and provide mechanical support. In this thesis, I devised the first whole cell model of the peptidoglycan by combining pieces of experimental data that are seemingly disparate. The model is “coarse-grained”, which means the atoms that make up the peptidoglycan are abstracted into beads that represent multiple atoms, which simplifies the calculations. Then I developed a “forcefield”, a complex set of energy functions, to conduct a dynamic simulation of the peptidoglycan. This simulation captured various experimentally determined physical properties of the peptidoglycan, demonstrating that the model is plausible. I then modeled the periplasm, which is the space the peptidoglycan resides in filled with proteins, at a whole cell scale for the first time. This model showed that some proteins can move through the peptidoglycan. One of the key technologies behind the periplasmic protein modeling was DeepMind’s AlphaFold2, which predicts the 3D structure of proteins with high accuracy. The second system that I modelled involved identifying where ligand binding sites can be placed inside hundreds of thousands of AlphaFold2 predicted protein structures. I showed that state-of-the-art techniques in protein structure prediction accurately reserve space for known and potentially novel binding sites.