Partnership


What did we do?


Description

The Edinburgh-UHAS_Ghana 2022 iGEM Team formed a fruitful partnership with Team Munich, where we held weekly meetings since July, devised a bioinformatic pipeline for the automated docking, mutation, and redocking of Transcriptional repressors to their ligands. With the skills gained in this partnership, we were also able to do docking simulations of Metallothioneins, which shaped the way we interpreted the results of Metallothionein directed evolution and hydrogel immobilization. Transcription Factor (TF) modelling gave us information on the potential success of our biosensors, and helped validate that a novel TF submitted to the parts repository could in fact be used in a biosensor the way we intended.

Results of the docking simulations can be found on the modelling page.

Timeline

July

In July, our iGEM Teams met at the European Meetup in Hamburg. Our projects are fundamentally different, with our team exploring cell free systems and dealing exclusively with prokaryotes, while Team Munich’s project is focussed around T-cells and Chimeric Antigen Receptors, but we found commonality in using transcriptional repressors to control our constructs. Because of this, we decided to pursue a purely bioinformatic collaboration, which soon evolved into a partnership, on structural and kinetic modelling of TFs.

Group photo of all iGEM teams at the European Meetup. Picture taken in DESY Hamburg.

Advisors on both our side and the Munich side of the partnership guided our choices in how to do this. First, we would need to get TF structures, perform energy minimisation and then docking simulations to see TF-ligand interactions. We were also interested in TF-DNA interactions, as well as in silico mutation as a way of improving, or learning more about our TFs. With this in mind, we began researching the available tools to perform this type of docking. We compiled a list of the available tools and how they could be used.

August

Come August, we had generated a list of available tools, and began testing out which tools could be most suited for our project, and which ones would not work. We had been looking into deep learning-based tools for binding motif prediction, as the binding motif for several of our TFs, as well as Team Munich’s TF were not known. Unfortunately, all the tools we had researched depend on the same ChIP-Seq dataset on the University of Toronto website, which was not available at the time. Therefore, we weren’t able to do any binding motif prediction, and started focussing more on TF-ligand interactions. The focus of August was structural prediction, as that must come first chronologically before mutation. We used AlphaFold to predict novel structures of uncharacterized TFs, dimerized them, and successfully performed the first docking simulations with a variety of tools. This was when we also decided to use AutoDock 4.2 for all our docking simulations, as it allowed for more customisation in docking, for example using covalently bonded docking instead of hydrogen bonds, as well as adding new atom types, which was a necessity to dock heavy metals. We began work on docking simulation automation, but due to their computationally strenuous nature, it took a long time for simulations to run and hence refine our docking parameters. We were also still exploring the avenue of DNA docking, and tested out the command line based tools mentioned before, but this yielded poor structures or results, or often failed altogether. The HDOCK server on the other hand, yielded high quality DNA docked structures. With these preliminary results, we were ready to go into September, where we would properly define our Bioinformatic pipeline for automated docking and mutation.

September

We had now decided to use AlphaFold for de novo structure prediction, AutoDock4.2 (with AutoGrid and MGLTools) for docking, the YASARA server for energy minimisation, the HDOCK Server for DNA docking, and a purpose built Jupyter notebook for mutation. The main focus of this month was mutation and automation, so that we could rapidly mutate and redock TFs, and see how mutations in the binding site changed kinetics of binding. Our team worked mainly on automation using the command line and bash scripts, while the Munich side focussed on the mutation protocol. By the end of September, we were able to rapidly generate mutants, and rapidly dock against libraries of PDB files. This meant that by the beginning of October we had generated all our final docking results and could begin analysing data.

October

Just because we had completed docking simulations, didn’t mean we necessarily had to end our partnership. In the weeks leading up to wiki freeze, we gave each other feedback on completed wiki pages, focussing on Human Practises and Education. We also decided that having spent so long on docking simulations, it would be useful to provide a tutorial on the basics of molecular docking, so that future iGEM teams have a stronger starting point approaching this potentially unfamiliar and technical concept.


How did both Teams Benefit?

Without partnership, this task would have been near impossible in the time frame, along with all the other things one must do for an iGEM project. We both consulted experts in our respective institutions, and were able to get guidance on various aspects of the docking simulations quite rapidly, which meant we could also start testing out different programs quickly. The advisors for both our teams were also integral in ensuring an accurate pipeline was established. Til from the Munich Team had also attended a seminar on rational enzyme design, where they discussed some docking simulations as well, which gave us a good framework of the order to run programs in. Arin Wongprommoon, an advisor on our side of the partnership, was able to help us extensively when troubleshooting AlphaFold. When dealing with automation of docking and mutation, our skillsets complemented each other. Maarten from our team dealt mainly with automation using the command line, while Lilly from the Munich Team mainly designed the Jupyter notebook we used for mutation, and it is likely that without these complementing skillsets we would not have been able to make such a successful pipeline. Finally, both teams were able to benefit even more with constructive feedback on wikis. We feel this partnership was very valuable and provided a lot of insight into both of our projects, and would encourage future teams to pursue bioinformatic partnerships as well.

Some features that took our teamwork beyond a collaboration include:


Docking Simulation Tutorial



Access Wiki Tools View Source Code