From zero to hero: raw data to mapped networks

Mapping

Use the Network >> mapping module to create features to modify the network map bode and edge properties.

The following is an example of how to format and combine analysis for network mapping. Check out the user manual for more details about this module’s features.

Methods

Use the `transformation` `map fold change` to format fold-change values (e.g. `FC_` results from the statistics module). This will create separate features for `FC_` magnitude(log2) and direction of change.

Select `calculate` and `save` to add the created features to the global data set.

Use a similar procedure to format `p-values` and combine features.

Use `map p-value` to evaluate an `inequqlity` comparing a variable to a `threshold` value.

Use `combine columns` to concatenate variables multiple variables into a single feature. Note, save the created variables from `map fold change` and `map p-value` `transformations` before using in the `combine columns` `transformation`.

Select `calculate` to view the results then use `save` to add the created features to the global data set.

Correlation analysis

Use the Network >> correlation module to calculate regularized correlations among all variables.

First, define the correlation `method`and `p-value` significance threshold and then select `calculate`.

Next, use regularize to prune the network to maintain the most confident relationships. Use `optimization` to define the initial method (`ric` or `stars` Jiang et al. (2019)). Next, you can select the `manual` optimization and select a custom regularization `lambda` (the higher the more strict the selection).

Select `calculate` and view the method results.

Explore and plot

For example the following mapped network displays positive and negative regularized correlations and the following node properties:

- size = magnitude of fold-change

- color and shape = direction of fold change and if the statistical p-value met the threshold criteria set in mapping.

The edges have been filtered to retain Tanimoto similarity > 0.8. See the user manual for more details about how to customize the visualization.

Use save to add the calculate network as an option to integrate with other available networks for the data set in network >> enrich.

Biochemical network

Use the Network >> biochemical module to calculate biochemical (substrate to product) relationships for all `KEGG` ids for metabolite measurements.

Define which variable in the data corresponds to the `KEGG` id or first use the network >> translate to translate it from metabolite names or other identifiers.

Next, `visualize`, create a `report` and save the calculated network.

Structural similarity network

Use the Network >> structural module to calculate molecule structural similarity relationships for all `PubChem CID` ids for metabolite measurements.

Define which variable in the data corresponds to the `CID` id or first use the network >> translate to translate it from metabolite names or other identifiers.

Next, `visualize`, create a `report` and save the calculated network.

Network enrichment

Use the Network >> enrich module to combine, update and visualize the calculated networks.

Select networks which should be combined into a single network. Note, selecting `single edges` removes any duplicated connections keeping the first observed instance as defined by the order of the `selected networks`.

`Visualize`, create a `report` and save the calculated network.

For example, the following `interactive` network displays the combined `biochemical`, `correlation` and `structural similarity` connections. The node size shows the magnitude of the fold change between `classes`. Node color encodes the direction and significance of the fold changes.

You can zoom and pan in the network and use the legend to filter the visualized nodes and edges. For example, the following subnetwork shows nodes for all significantly altered comparisons.

Save and export

Save the created network to make it available for export.

To continue optimizing the network using external tools, use the data module to export (save) the selected network `nodes` and `edges`.

Grapov, D., K. Wanichthanarak, and O. Fiehn. 2015. “MetaMapR: Pathway Independent Metabolomic Network Analysis Incorporating Unknowns.” Journal Article. Bioinformatics 31 (16): 2757–60. https://doi.org/10.1093/bioinformatics/btv194.

Jiang, Haoming, Xinyu Fei, Han Liu, Kathryn Roeder, John Lafferty, Larry Wasserman, Xingguo Li, and Tuo Zhao. 2019. Huge: High-Dimensional Undirected Graph Estimation. https://CRAN.R-project.org/package=huge.

Mapping

Use the Network >> mapping module to create features to modify the network map bode and edge properties.

The following is an example of how to format and combine analysis for network mapping. Check out the user manual for more details about this module’s features.

Methods

Use the transformation map fold change to format fold-change values (e.g. FC_ results from the statistics module). This will create separate features for FC_ magnitude(log2) and direction of change.

Select calculate and save to add the created features to the global data set.

Use a similar procedure to format p-values and combine features.

Use map p-value to evaluate an inequqlity comparing a variable to a threshold value.

Use combine columns to concatenate variables multiple variables into a single feature. Note, save the created variables from map fold change and map p-value transformations before using in the combine columns transformation.

Select calculate to view the results then use save to add the created features to the global data set.

Correlation analysis

Use the Network >> correlation module to calculate regularized correlations among all variables.

First, define the correlation methodand p-value significance threshold and then select calculate.

Select calculate and view the method results.

Explore and plot

Use the Network >> visualize menu to create mapped networks integrated with the calculated relationships.

For example the following mapped network displays positive and negative regularized correlations and the following node properties:

- size = magnitude of fold-change

- color and shape = direction of fold change and if the statistical p-value met the threshold criteria set in mapping.

The edges have been filtered to retain Tanimoto similarity > 0.8. See the user manual for more details about how to customize the visualization.

Use save to add the calculate network as an option to integrate with other available networks for the data set in network >> enrich.

Biochemical network

Use the Network >> biochemical module to calculate biochemical (substrate to product) relationships for all KEGG ids for metabolite measurements.

Define which variable in the data corresponds to the KEGG id or first use the network >> translate to translate it from metabolite names or other identifiers.

Next, visualize, create a report and save the calculated network.

Structural similarity network

Use the Network >> structural module to calculate molecule structural similarity relationships for all PubChem CID ids for metabolite measurements.

Define which variable in the data corresponds to the CID id or first use the network >> translate to translate it from metabolite names or other identifiers.

Next, visualize, create a report and save the calculated network.

Network enrichment

Use the Network >> enrich module to combine, update and visualize the calculated networks.

Select networks which should be combined into a single network. Note, selecting single edges removes any duplicated connections keeping the first observed instance as defined by the order of the selected networks.

The update menu can be used to update the combined networks variable meta data (_col_meta).

Visualize, create a report and save the calculated network.

Use the nodes menu to specify node labels, color, size and shape.

Use the edges menu to filter edges. For example, filter structural similarity edges to keep only those showing Tanimoto similarity greater than 0.8 Grapov, Wanichthanarak, and Fiehn (2015).

For example, the following interactive network displays the combined biochemical, correlation and structural similarity connections. The node size shows the magnitude of the fold change between classes. Node color encodes the direction and significance of the fold changes.

You can zoom and pan in the network and use the legend to filter the visualized nodes and edges. For example, the following subnetwork shows nodes for all significantly altered comparisons.

Save and export

Save the created network to make it available for export.

To continue optimizing the network using external tools, use the data module to export (save) the selected network nodes and edges.

Use the `transformation` `map fold change` to format fold-change values (e.g. `FC_` results from the statistics module). This will create separate features for `FC_` magnitude(log2) and direction of change.

Select `calculate` and `save` to add the created features to the global data set.

Use a similar procedure to format `p-values` and combine features.

Use `map p-value` to evaluate an `inequqlity` comparing a variable to a `threshold` value.

Use `combine columns` to concatenate variables multiple variables into a single feature. Note, save the created variables from `map fold change` and `map p-value` `transformations` before using in the `combine columns` `transformation`.

Select `calculate` to view the results then use `save` to add the created features to the global data set.

First, define the correlation `method`and `p-value` significance threshold and then select `calculate`.

Select `calculate` and view the method results.

Use the Network >> biochemical module to calculate biochemical (substrate to product) relationships for all `KEGG` ids for metabolite measurements.

Define which variable in the data corresponds to the `KEGG` id or first use the network >> translate to translate it from metabolite names or other identifiers.

Next, `visualize`, create a `report` and save the calculated network.

Use the Network >> structural module to calculate molecule structural similarity relationships for all `PubChem CID` ids for metabolite measurements.

Define which variable in the data corresponds to the `CID` id or first use the network >> translate to translate it from metabolite names or other identifiers.

Next, `visualize`, create a `report` and save the calculated network.

Select networks which should be combined into a single network. Note, selecting `single edges` removes any duplicated connections keeping the first observed instance as defined by the order of the `selected networks`.

The update menu can be used to update the combined networks variable meta data (`_col_meta`).

`Visualize`, create a `report` and save the calculated network.

Use the edges menu to filter edges. For example, filter `structural similarity` edges to keep only those showing Tanimoto similarity `greater than` 0.8 Grapov, Wanichthanarak, and Fiehn (2015).

For example, the following `interactive` network displays the combined `biochemical`, `correlation` and `structural similarity` connections. The node size shows the magnitude of the fold change between `classes`. Node color encodes the direction and significance of the fold changes.

To continue optimizing the network using external tools, use the data module to export (save) the selected network `nodes` and `edges`.