Identifying a Novel degradation Pathway with KBase discovery pipeline and PDB tools¶

Narrative synopsis¶

• In this Narrative workflow, we demonstrate the KBase discovery pipeline in identifying potential gene candidates on a novel Pyridine degradation pathway in Micrococcus luteus. Here (i) we use cheminformatics analysis to propose new biochemistry then, (ii) we use metabolic modeling and omics data to identify potential gene candidates, (iii)we then query the PDB to fetch metadata/annotations for experimentally resolved structures corresponding to the gene candidates. For these selected structures, later in this workshop, PDB team will demonstrate (iv) deriving co-crystallized structures with the substrates of interest that bolster the confidence of the identified gene candidates on this novel degradation pathway. Finally, the gene candidates can be experimentally verified.

1. ) Annotate Mluteus genome using the RAST, Prokka and DRAM annotation pipelines 1a.) Explore the annotated genome
2. ) Construct a Draft Metabolic Model/Metabolic network based on the functional annotations
3. ) Generate network of hypothetical degradation reactions based on pyridine with Pickaxe
4. ) Creating a Base Media for Gapfilling
5. ) Filling knowledge gaps in the metabolic network - Gapfilling Metabolic Model 5a.) Creating a Pyridine Minimal Media for model simulation/ run Flux Balance Analysis (FBA)
6. ) Running FBA on M.luteus against Pyridine Minimal Media aerobically
7. ) Visualizing Noval Pyrdine degredation pathway and the fluxes in an Escher map
8. ) Find potential gene candidates - Use differential expression analysis and gene clustering data to fliter highly expressed genes relavant to pyridine degredation 9.) Use of PDB structural evidence in identifying key steps of the pyridine degradation pathway 10.) Further investigate experimental structures that corresponds to candidate genes

Next, we follow another interesting example in the Arabidopsis riboflavin pathway on this narrative workflow that shows the value of applying computational tools in KBase and PDB to address important scientific questions.</p>

1. Annotate Mluteus genome using the RAST, Prokka and DRAM annotation pipelines¶

Here we annotate the Mluteus genome using three annotation pipelines which derive functional annoations for each gene in the genome. We annotate with three seperate algorithms which increase the chances of assigning functions for maximum number of genes in the genome

1a. Explore the Annotated Genome¶

Below, the M.luteus genome is shown in a genome viewer. This viewer provides a concise, text-based overview of the genome as well as its contigs and genes.

In the Contigs and Features tabs, each entry is clickable, opening either a browser for the contig or another tab with expanded information about the gene. You can sort these entries by clicking on a column header to sort by that field (e.g., Length). Clicking the same column header again will reverse the sort order.

The M.luteus genome has two contigs. Click on one to see neighboring genes and potential operons in this species.

To further explore this genome, click on "Browse Features" tab, where you can search for gene annotations/functions by name (e.g; pyruvate synthase, EC numbers etc.), extract DNA or protein sequences, explore the neigboring genes/gene clusters

2. Construct a Draft Metabolic Model/Metabolic network based on the functional annotations¶

We use our M.luteus genome previoulsy assembled and annotated with multiple annotation algorithms for draft metabolic modeling reconstruction.

Reference tutorial narrative on Metabolic Model Construction¶

The output (above) of the Build Metabolic Model app shows information about the resulting gapfilled model. (Note that although the object type is “FBA Model,” we have not actually performed a flux balance analysis yet.)

There are eight tabs for browsing the data in the model: Overview, Reactions, Compounds, Genes, Compartments, Biomass, Gapfilling, and Pathways.

• Overview — Summary of key information about the model, including the associated genome, number of reactions, and number of compounds.
• Reactions — Detailed reaction information, including reaction ID, name, biochemical equation, the associated gene IDs, and whether or not the reaction was added by the gapfilling stage.
• Compounds — Information about compounds in the model, including chemical formula and charge.
• Genes — Gene IDs and associated reaction IDs.
• Compartments — Subcellular localization of the compounds and enzymes. Typically, there are three types of compartments in microbes: Cytosol (c), -Periplasm (p), and Extracellular (e). Reactions and compounds belonging to each compartment are identified using compartment notation (e.g., rxn00001[c0], cpd00001[c0]). The integer associated with the compartment (e.g., the 0 in c0) represents the index number of the model. For a single-species model, this number will always be zero, but if individual models are merged into a community model, each sub-model will then be assigned a distinct index.
• Biomass — Biomass composition of the model. Typically biomass is represented in the model as an equation where biomass compounds and ATP would make 1 gram of biomass. After clicking on the Biomass tab, the coefficients of each biomass component are listed in the Coefficient column. Negative coefficients represent the compounds on the left side of the biomass equation, and positive coefficients represent the compounds on the right side of the equation.
• Gapfilling — Reactions that were added to fill metabolic gaps resulting from missing or inconsistent annotations. During the gapfilling process, an optimization algorithm adds a minimal number of reactions and compounds to make the biochemical network generate biomass. Currently, this tab does not show anything because gapfilling indiciation was moved to the Reactions tab.
• Pathways — KEGG maps that represent the metabolic network of the model. Click on the name of a map (e.g., TCA cycle) to see the presence or absence of the reactions (blue).

3. Generate network of hypothetical degradation reactions based on Pyridine with Pickaxe¶

To generate some potential utilization routes for pyridine, we use Pickaxe app. This tool uses a set of general reaction rules which are curated from known biochemistry as the figure below demonstrates. These reactions can be applied to novel substrates like pyridine to propose new chemical transformations.

4. Creating a Base Media for Gapfilling¶

You can construct any custom media with the Edit Media app. Here we use an existing media formulation (e.g; Glucose Minimal Media) that we can copy from our reference media and remove Glucose,the sole carbon source from the media creating a base media base media that have all necessary salts, Oxygen, Nitrogen, Sulfur, Phosphate except for the carbon source. We discuss having Pyridine as the carbon source in the following step of Gapfill Metabolic Model step (Filling knowledge gaps in the metabolic network - Gapfilling Metabolic Model).

5. Filling knowledge gaps in the metabolic network - Gapfilling Metabolic Model¶

Typically, draft metabolic models tend to have metabolic gaps due to missing or incomplete annotations. In this workflow, the metaboic gap that we are interested is the Pyridine degredation as the pathway is not chracterized. We have used the PickAxe app above to generate potential noval reactions and pathways for pyridine degration. Now, in next step we use the PickAxe output (Selected under Source Gapfill Model) to fill the pyridine degredation gap in the M.luteus metabolic model.

As for the Media, we use a base media that have all necessary salts, Oxygen, Nitrogen, Sulfur, Phosphate - (Selected under Media) and the sole carbon source Pyridine will be selected under Source model media supplement option.

There are eight tabs for browsing the data (above) in the model: Overview, Reactions, Compounds, Genes, Compartments, Biomass, Gapfilling, and Pathways.

• Overview — Summary of key information about the model, including the associated genome, number of reactions, and number of compounds.
• Reactions — Detailed reaction information, including reaction ID, name, biochemical equation, the associated gene IDs, and whether or not the reaction was added by the gapfilling stage.
• Compounds — Information about compounds in the model, including chemical formula and charge.
• Genes — Gene IDs and associated reaction IDs. In a typical genome-scale model output table genes tab is populated, however, for metagenome models, given that there are extremly large number of genes affecting the efficient loading and browsing of the table, we do not display the genes.
• Compartments — Subcellular localization of the compounds and enzymes. Typically, there are three types of compartments in microbes: Cytosol (c), -Periplasm (p), and Extracellular (e). Reactions and compounds belonging to each compartment are identified using compartment notation (e.g., rxn00001[c0], cpd00001[c0]). The integer associated with the compartment (e.g., the 0 in c0) represents the index number of the model. For a single-species model, this number will always be zero, but if individual models are merged into a community model, each sub-model will then be assigned a distinct index.
• Biomass — Biomass composition of the model. Typically biomass is represented in the model as an equation where biomass compounds and ATP would make 1 gram of biomass. After clicking on the Biomass tab, the coefficients of each biomass component are listed in the Coefficient column. Negative coefficients represent the compounds on the left side of the biomass equation, and positive coefficients represent the compounds on the right side of the equation.
• Gapfilling — Reactions that were added to fill metabolic gaps resulting from missing or inconsistent annotations. During the gapfilling process, an optimization algorithm adds a minimal number of reactions and compounds to make the biochemical network generate biomass. Currently, this tab does not show anything because gapfilling indiciation was moved to the Reactions tab. Pathways — KEGG maps that represent the metabolic network of the model. Click on the name of a map (e.g., TCA cycle) to see the presence or absence of the reactions (blue).

5a. Creating a Pyridine Minimal Media for model simulation/ run Flux Balance Analysis (FBA)¶

In order to simulate metabolic moodels (to run FBA), we need a media formulation. In this work flow we use the custom media formulation Pyridine Minimal media. You can construct any custom media with the app Edit Media. Here we use an existing media formulation (e.g; Glucose Minimal Media) that we can copy from our reference media and replace with Pyridine creating Pyridine minimal media.

6. Running FBA on M.luteus against Pyridine Minimal Media aerobically¶

Now we are going to run flux balance analysis on one of the bins (genome-scale model), which will simulate how this would grow on Pyridine-minimal media.

Useful articles on Flux Balance Analysis and Flux Variability Analysis¶

What is FBA - https://www.ncbi.nlm.nih.gov/pmc/articles/PMC3108565/

Flux Variability Analysis (FVA) - https://www.ncbi.nlm.nih.gov/pubmed/14642354

Reaction classifiers are assigned when using Flux Variability Analysis, FVA. In FVA, the global objective (biomass) is fixed at its optimal value, then each reaction, iteratively, is optimized independently to find both the maximal and minimal value that is possible given that the global objective must still be reached. FVA analysis composed into four categories in the following FBA. (See the column "class" in the FBA output data)

• Variable – the reaction has positive maximal and negative minimal values, meaning that it can go in either direction.
• Positive variable – the reaction has a positive maximal, and a zero minimal, meaning that it can either be zero, or it can go from left to right.
• Negative variable – the reaction has a zero maximal, and a negative minimal, meaning it can either be zero, or it can go from right to left.
• Blocked – the reaction is blocked and cannot have a non-zero value.

When the FBA analysis finishes, information on the flux distribution is displayed in a table with six tabs: Overview, Reaction fluxes, Exchange fluxes, Genes, Biomass, and Pathways (see above).

• Overview — Among the summary information in this tab is the objective value (growth of the model), which is important because it represents the maximum achievable flux through the biomass reaction of the metabolic model. An objective value of 0 or a value very close to 0 means that the model did not grow on the specified media. This tab also lists other information, including the genome, media formulation, number of reactions, and number of compounds associated with the FBA.
• Reaction fluxes — Numerical flux values, minimum and maximum flux bounds, biochemical equations, and associated genes for each reaction in the model. This information represents the fluxes through all internal reactions that allow for growth and byproduct creation. These fluxes can be further broken down into biological pathways of interest (see Pathways tab). A user may ask, for example, “what compounds consumed or excreted?” or “What are the high flux reactions or pathways?”
• Exchange fluxes — These fluxes describe the rates at which nutrients are taken in and byproducts are secreted. Positive exchange flux values represent the uptake of compounds, and negative exchange flux values represent the excretion of compounds.

• Genes — This tab displays the gene knockout information, if any.

• Biomass — Biomass composition of the model is displayed. Typically, biomass is represented in the model as an equation where biomass compounds and ATP would make 1 gram of biomass. After clicking on the Biomass tab, the coefficients of each biomass component are listed in the Coefficient column. Negative coefficients represent the compounds on the left side of the biomass equation, and positive coefficients represent the compounds on the right side of the equation.

• Pathways — This tab displays KEGG maps that represent the metabolic network of the model. Click on the name of a map (e.g., TCA cycle) to see the presence or absence of reactions (blue) and fluxes (positive fluxes are shades of red; negative fluxes are shades of green).

7. Visualizing Pyrdine degredation pathway and the fluxes in an Escher map¶

Identifying potential gene candidates for the pyridine degradation¶

8. Use differential expression analysis (Glucose vs Pyridine) and gene clustering data to fliter highly expressed genes relevant to pyridine degredation¶

Now we have demonstrated a potential novel pathway for pyridine degredation, next, we can work on identifying the potential gene candidates. From the gapfilling and FBA steps, we can see the novel pyridine degradation reactions are associated with partial EC number 1.14.13 -. In our genome, there are about 30 genes are assigned with the first three digits of the EC number 1.14.13. We use the (i) differential expression data to filter out highly expressed genes (green) and (ii) the gene clustering data. Differential Expression analysis Narrative can be found here.

9. Use of PDB structural evidence in identifying key steps of the pyridine degradation pathway¶

The PDB annotation app (below) fetch any available structural data evidence in PDB that are homologous to genes in the M.luteus genome¶

While the highly expressed genes with EC 1.14.13 narrow down the list of gene candidates, the gene clustering data/neighboring genes in the same operon provide valuable insights on key enzymatic steps of the degradation pathway. By surveying the gene cluster/neibouring genes with the MLuteus_masurca_RAST.CDS.3484 against PDB structural evidence, we can find the MLuteus_masurca_RAST.CDS.3483 gene, a phenylacetate dehydrogenase (paaZ) linked to a literature explaining the key step of ring opening enzyme on phenylacetate, a substrate that is chemically similar to pyridine.

10. Further investigate experimental structures that corresponds to candidate genes¶

Here we can query experimentally resolved structures that are corresponding to the potential gene candidates

Tune in for:

• Query and learn from co-crystalized structures with the docking of the substrate

• Align experimental and computational structures to aid binding site identification and functional characterization