TL;DR
This paper introduces EMMA, a computational workflow that leverages enzyme promiscuity to expand E. coli's metabolic model (iML1515). By predicting novel reactions and metabolites, EMMA aims to create a more complete picture of cellular metabolism, which is crucial for advancing systems biology and metabolic engineering.
What Problem Does This Paper Address?
Metabolic models are powerful tools for understanding and engineering cellular processes. However, these models are often incomplete because not all enzymatic activities are fully known or annotated. This incompleteness can lead to suboptimal predictions and unexpected experimental outcomes. The authors propose that a significant source of this "missing" metabolism is enzyme promiscuity – the ability of enzymes to catalyze reactions beyond their primary, annotated function. While enzyme promiscuity is widely accepted, systematic analyses to predict its impact on genome-scale metabolic models have been lacking. This paper aims to fill that gap by developing a method to predict and incorporate these promiscuous reactions and their resulting metabolites into existing models.
Methods
The core of the EMMA workflow relies on a tool called PROXIMAL. PROXIMAL works by:
- Identifying Reaction Patterns: It analyzes known biochemical reactions from the KEGG database to identify "Reaction Center, Difference Region, and Matched Region" (RDM) patterns. These patterns capture the structural transformations that occur during a reaction.
- Generating Biotransformation Operators: Using the RDM patterns associated with enzymes in the E. coli iML1515 model, PROXIMAL generates operators that represent potential biotransformations.
- Predicting Promiscuous Products: These operators are then applied to the known metabolites within the iML1515 model to predict new, putative metabolites that could be formed through promiscuous enzymatic activity.
The EMMA workflow then proceeds in three steps:
- Step 1: Predict Promiscuous Products: EMMA uses PROXIMAL to generate a list of potential new metabolites.
- Step 2: Compare with Metabolomics Data: These predicted metabolites are compared against the Escherichia coli Metabolome Database (ECMDB) to identify "verifiable" predicted promiscuous transformations.
- Step 3: Curation of Stoichiometric Reactions: For verifiable transformations, the authors manually examine the reactions, ensure they are stoichiometrically balanced by adding cofactors, and then compare them against existing databases like EcoCyc and KEGG. This categorization helps understand the novelty and origin of the predicted reactions.
Datasets
The study utilizes the following datasets:
- E. coli iML1515 metabolic model: This is the primary model used as the basis for predictions.
- Escherichia coli Metabolome Database (ECMDB): This database, containing 3,760 metabolites detected in E. coli, is used to validate predicted metabolites.
- KEGG (Kyoto Encyclopedia of Genes and Genomes): Used for its comprehensive database of biochemical reactions and pathways.
- EcoCyc: Another database of E. coli metabolic information, used for comparison and validation.
Key Insights and Ideas
- Enzyme promiscuity is a significant, underutilized source of metabolic novelty: The study demonstrates that enzymes can catalyze many more reactions than currently documented, leading to the formation of new metabolites.
- A systematic computational approach can uncover these "hidden" metabolic pathways: EMMA, powered by PROXIMAL, provides a framework to move beyond manual curation and leverage computational power to explore enzyme promiscuity.
- Bridging computational predictions with experimental metabolomics is key: Validating predicted metabolites against real-world metabolomics data (ECMDB) is crucial for identifying biologically relevant transformations.
- Categorizing novel reactions helps prioritize future research: The classification of predicted reactions into categories (C1-C4) based on their presence in existing databases provides a roadmap for understanding their novelty and potential impact.
Key Results
The EMMA workflow was applied to the E. coli iML1515 model, resulting in:
- 1,875 biotransformation operators generated by PROXIMAL.
- 1,423 predicted metabolites from 106 high-concentration metabolites.
- 16 new metabolites and 23 new reactions recommended for augmenting the iML1515 model.
- Of these 23 reactions:
- 4 were classified as Category 1 (C1): Novel transformations parallel to existing ones, catalyzed by different enzymes.
- 10 were classified as Category 2 (C2): Reactions known in E. coli but missing from iML1515 (curation issues).
- 5 were classified as Category 3 (C3): Reactions known in other organisms but not yet documented in E. coli.
- 4 were classified as Category 4 (C4): Novel reactions not cataloged in KEGG for any organism.
- Notably, 9 of the 16 predicted metabolites were found in ECMDB but not in other E. coli databases, highlighting their novelty.
Limitations and Future Work
The authors acknowledge several limitations and suggest future directions:
- Computational limitations: PROXIMAL did not consider if predicted products of promiscuous reactions could themselves act as new substrates for further promiscuous reactions, due to the large number of potential products.
- Experimental validation needed: While the study validates predictions against metabolomics data, experimental validation of the novel reactions and metabolites is the next crucial step.
- Improving prediction accuracy: The authors are developing machine learning techniques to enhance PROXIMAL's prediction accuracy.
- Expanding PROXIMAL's capabilities: Future work could involve upgrading PROXIMAL to consider enzymes that act on multiple reaction centers within a metabolite.
- Broader applicability: The workflow is designed to be applicable to any organism with a metabolic model, promising enhanced understanding of diverse metabolic systems.
Summary updated