Team:Manchester/Pathway Design
Pathway Design
- We successfully applied innovative methods of retrosynthesis to design a new pathway towards a compound for which the natural biosynthetic route is unknown.
- We used computational tools to select enzymes that are predicted to catalyse the novel reactions in our predicted pathway.
- We prioritized the selected enzymes to identify the most promising candidates for experimental implementation in engineered bacteria.
What is Pathway Design?
Metabolic pathway design is an emerging discipline at the core of synthetic biology. Using an engineering perspective, it opens the door to uncharted possibilities in biology – the possibility of creating artificial metabolic pathways to mass-produce new-to-nature compounds using computer aided design tools. Synthetic biology requires a multidisciplinary approach encompassing knowledge of: biochemistry; cell biology; systems biology; biotechnology; bioinformatics; and model simulation and optimization, among others.
Bio-manufacturing and a new bioeconomy are on the rise as new models of production are both increasingly sustainable and affordable. Pathway design will be a key component of this revolution as the demand for chemicals increases. The discipline itself and our approach to it follows the Design–Build–Test–Learn (DBTL) Cycle that defines the current approach to synthetic biology – and the approaches of both iGEM and the Manchester Institute of Biotechnology, the institution that hosts our team.
This step of the project is part of the Design component of the Synthetic biology DBTL Cycle; however, the DBTL philosophy was employed on a small scale at each step in the process.
The first step was to choose a compound of interest, in our case Hipposudoric Acid (HA). We performed a wide literature search to make sure the compound had not been produced in bacteria before. In our search, we found several key articles which would play an important role in our project, the most important of which for our pathway design was the paper by Hashimoto et al. published in 2007, where the authors discussed a possible biological precursor of HA.
Hipposudoric acid is a natural compound, but its natural biosynthetic route in the hippopotamus is unknown. Thus, the next step of our project used retrosynthesis to come up with a potential biosynthetic route that we could implement in engineered bacteria.
Figure 1. Overview of the Pathway Design workflow. Each of the major steps going form Hippopotamus to Hipposudoric Acid highlighting each of the associated computational tools.
Pathway Design — RetroPath2.0 & rp2paths
Retropath2.0 is a retrosynthetic tool, and the first step of our metabolic design pipeline. In short, the software takes our compound as input and tries to find routes from this compound to any metabolite present in the natural metabolism of our host system, E. coli (2). The software tries to trace reactions from our compound, HA, to possible precursors, i.e. compounds that in a single reaction step could be transformed into our target. Then, it performs the same operation with this first batch of possible precursors, and then iterates the same operation several times, until it reaches a metabolite present in the E.coli central metabolism.
Working with RetroPath2.0 for the first time required a lot of troubleshooting, and it could not find a direct precursor to HA. This is where the literature was key: we found a likely precursor, Homogentisic Acid (HGA), which is non-enzymatically converted to HA (which was the reason that RetroPath2.0 didn’t identify the reaction. When we used RetroPath2.0 to find routes to HGA, we obtained more favorable results.(1) The final reaction from HGA to HA was considered manually using chemical analysis.
The output from RetroPath2.0 is a series of reactions connecting compounds independently. The next step, a crucial one, is organizing these reactions into pathways and validating them.
Figure 2. RetroPath2.0 on the screen and rp2path output pathways examples.
Rp2paths takes the output from RetroPath2.0 and assembles pathways by connecting the reactions found by the first tool. The final output from the retrosynthesis and the pathway assembly was a collection of seven possible pathways, all deriving the Homogentisic Acid from Tyrosine derivatives in E. coli metabolism.
Enzyme Selection — Selenzyme
Choosing the correct enzymes to catalyze the reactions in this novel pathway is just as important as determining the pathway itself. Selenzyme is an enzyme prediction tool that assigns enzymes to the reactions assembled into pathways by rp2paths and finds the enzymes best able to catalyze these reactions.(3)
Selenzyme uses the information from its database to recommend enzymes to perform the predicted reactions, as well as calculating a set of scores for each enzyme indicating various confidence metrics. These scores are then combined to rank the proposed enzymes for each reaction in the pathway according to their expected ability to catalyze our target reactions.
The most important scores are rsim_score and ec_match_score which rank the enzymes based on the similarity between the reaction they normally catalyze and the reaction of our novel pathway in the output from rp2paths.
Enzymes which have been experimentally characterized will be ranked higher than those for which only DNA sequence information is available; is is reflected by the protein_evidence_penalty metric. Additionally, the tool tries to find enzymes from organisms taxonomically close to the host organism selected, in our case E. coli K12; this is measured in the host_taxonomic_distance_penalty score. The last score is the rule_penalty a measure of the sequence diversity of the enzymes.
The final ranking is taken by combining the aforementioned scores into the following algorithm, the default method for organizing the enzymes in Selenzyme.
sequence_score = +100 J +1 I -1 E -0.1 H
- J = rsim_score
- I = rule_penalty
- E = host_taxonomic_distance_penalty
- H = protein_evidence_penalty metric
The next steps will be Plasmid Design, and Flux Balance Analysis to simulate the best conditions for the production of the compound of interest.
Timeline for the computational work and troubleshooting
- 25th March – First download of KNIME and RetroPath2.0
- 30th March – First Troubleshooting meeting, installation problems and questions about tutorial data
- 3rd April – After training, first attempts to input HA
- 15th April – Meeting to discuss incomplete output and possible reasons for it
- 23rd April – Switch from GUI to command line version
- 6th May – Meeting for troubleshooting conda package errors and compatibility with KNIME
- 7th May – Results for HA retrosynthesis
- 9th May – Validation of results show they were a false positive
- 3rd June – After many fruitless attempts, switch from HA to HGA
- 9th June – Installation of different KNIME versions
- 20th June – Troubleshooting of conda environment
- 2nd July – First valid results
- 4th July – Meeting to discuss irreproducibility of results
- 29th July – Results replicated consistently
- 4th August – Validation and assembly with rp2paths
- 13th August – Selenzyme installation with Docker
- 17th August – Selenzyme troubleshooting
- 23rd August – Results from Selenzyme
