Inherent Dynamics Visualizer, an Interactive Application for Evaluating and Visualizing Outputs from a Gene Regulatory Network Inference Pipeline
December 7th, 2021
The Inherent Dynamics Visualizer is an interactive visualization package that connects to a gene regulatory network inference tool for enhanced, streamlined generation of functional network models. The visualizer can be used to make more informed decisions for parameterizing the inference tool, thus increasing confidence in the resulting models.
The Inherent Dynamics Visualizer makes working with several inference methods and understanding their results easier for researchers which in turn allows for the accelerated production of functional network models. The main advantage of the Inherent Dynamics Visualizer is visualizing and exploring intermediate results for informing parameters and input file contents of downstream steps. The Inherent Dynamics Visualizer facilitates exploration and understanding of results from several inference tools.
However, we recommend referencing each tool's documentation to better understand how parameter choices impact results. In a new IDP configuration file that parameterizes the node finding step, type data file equal to, annotation file equal to, output file equal to, number process equal to, and IDVconnection equal to true on individual lines. After the equal to sign for data file, type the path to and name of the respective time series file with a comma after the name.
For annotation file, type the path to and name of the annotation file. For output file, type the path to and name of the results folder. And for number process, type the number of processes the IDP should use.
In the same text file after the main arguments, type in the order presented DLxJTK arguments in square brackets periods equal to and DLxJTK cutoff equal to on individual lines. For periods, if one time series dataset is being used, type each period length separated by commas after the equal to sign. For more than one time series dataset, type each set of period lengths as before, but place square brackets around each set and place a comma between the sets.
For DLxJTK cutoff, after the equal to sign, type an integer specifying the maximum number of genes to retain in the gene list file output by De Lichtenburg by JTK-CYCLE. Next, run the IDP using the created configuration file by running the indicated command with the appropriate filename in the terminal. In the terminal, move to the directory named Inherent Dynamics Visualizer and enter the indicated command.
Then in a web browser, enter the URL shown. Next, click on the node finding tab and select the node finding folder of interest from the dropdown menu. To extend or shorten the gene list table, click on the up or down arrows or manually enter an integer between one and 50 in the box next to gene expression of DLxJTK ranked genes.
In the gene list table, click on the box beside a gene to view its gene expression profile and a line graph. Multiple genes can be added. Then download the gene list into the file format needed for the edge finding step by clicking on the download gene list button.
In the editable gene annotation table, label a gene as a target, a regulator, or both. If a gene is a regulator, label the gene as an activator, repressor, or both. Finally, click on the download annotation file button to download the annotation file into the file format needed for the edge finding step.
In a new IDP configuration file that parameterizes the edge finding step, for gene list file, enter the path to and name of the generated gene list file after the equal to sign. For edge score column, enter either PLD or norm loss to specify which dataframe column from the LEM py output is used to filter the edges. Next, select either edge score threshold or num edges for list and delete the other.
If edge score threshold was selected, enter a number between zero and one. If num edges for list was selected, enter a value equal to or less than the number of possible edges. Then select either seed threshold or num edges for seed and delete the other.
If seed threshold was selected, enter a number between zero and one. If num edges for seed was selected, enter a value equal to or less than the number of possible edges. Next, run the IDP using the created configuration file as demonstrated earlier.
Select or deselect edges from the edge table by clicking the corresponding checkboxes adjacent to each edge to add or remove edges from the seed network. Then click on the download DSGRN network specification button to download the seed network in the DSGRN network specification format. After selecting the edges to be included in the edge list file used in the network finding by clicking the corresponding checkboxes from the edge table, click on download node and edge lists to download the node list and edge list files in the format required for their use in network finding.
In a new IDP configuration file that parameterizes the network finding step, for seed net file, edge list file, and node list file, enter the path to and name of the seed network file and the edge and node list files. For range operations, type two numbers separated by a comma after the equal to sign. The first and second numbers are the minimum and the maximum number of addition or removal of nodes or edges per network made respectively.
For num neighbors, enter a number that represents how many networks to find in the network finding. And for max params, enter a number that represents the maximum number of DSGRN parameters to allow for a network. For add node, add edge, remove node, and remove edge, enter values between zero and one after the equal to sign.
The numbers must sum to one. Then run the IDP using the created configuration file as demonstrated earlier. To generate an edge prevalence table, select networks using the following two options.
For option one, input lower and upper bounds on query results by inputting minimum and maximum values in the input boxes corresponding to the x-axis and y-axis of the plot. For option two, click and drag over the scatterplot to draw a box around the networks to be included. After selection or entering of input bounds, click the get edge prevalence from selected networks button.
Next, input an integer in the network index input box to display a single network from the selection. Then click on download DSGRN network specification to download the display network in the DSGRN network specification format. Using the checkboxes corresponding to each edge, select edges to be included in the network or motif used for the similarity analysis.
Then click on submit to create the similarity scatterplot for the selected motif or network. To select a network or set of networks, click and drag over the scatterplot. Draw a box around the networks to be included to generate an edge prevalence table and to view the networks along with the respective query results.
Download the edge prevalence table by clicking download table. Then download the display network for similarity analysis by clicking on download DSGRN network specification as demonstrated earlier. This protocol was applied to the core oscillator gene regulatory network of the yeast cell cycle.
Results from running every step of the IDP consecutively without using the IDV between steps are shown here. This fully parameterized run of the IDP produced results for node and edge finding. Yet in network finding, no model admissible networks were discovered.
A set of known yeast cell cycle regulators was then selected from the Saccharomyces genome database and known regulatory relationships between genes were extracted from yeast tracked. The gene list table was extended to find the remaining gene in the gene regulatory network model and genes were deselected to remove genes not found in the same model. A new IDP configuration file was parameterized for the edge finding step with the new gene list and annotation file, and results were loaded in into the IDV.
Edges without experimental evidence were removed from the seed network. After creating a seed network well-supported by experimental evidence, 37 model admissible networks were found in the network finding step, of which 24 can stably oscillate. Of these 24 networks, the best performers were two networks that matched the data at 50%of their stably oscillating model parameters.
When the ability to remove edges during network generation was added, 612 networks were found with 67%of these networks having the capacity to oscillate stably. Interestingly, 82 networks capable of stable oscillatory dynamics were not capable of producing dynamics similar to those seen in the data. And of the 411 networks, 124 exhibited robust matches to the data.
Biologically feasible parameter space is unknown to DSGRN, but incorporating biological information in node and edge finding helps constrain network finding to biologically reasonable regions in the space of all networks. ODE modeling of the networks using parameters from the edge finding step can be performed to further test the functionality of the networks in silico.
View the full transcript and gain access to thousands of scientific videos
Overview
The Inherent Dynamics Visualizer is designed to enhance the usability of gene regulatory network inference methods, enabling researchers to create functional network models more efficiently. By providing tools to visualize and explore intermediate results, it aids in parameter selection and understanding, ultimately fostering confidence in the resultant models.
Key Study Components
Research Area
- Gene regulatory network inference
- Modeling complexity in biological systems
- Visualization tools in computational biology
Background
- Importance of parameterization in model accuracy
- Collaboration of visualization with inference methods
- Need for streamlined model development
Methods Used
- Interactive visualization software
- Gene regulatory networks
- Configuration file setups for various inference steps
Main Results
- Facilitated exploration of inference results
- Streamlined configuration for downstream modeling tasks
- Enhanced understanding of parameter impacts through visual tools
Conclusions
- The study demonstrates the efficacy of the Inherent Dynamics Visualizer in network model development.
- This tool significantly improves the researcher’s ability to derive dependable biological insights through better visualization and interaction with inferential tools.
Frequently Asked Questions
Executive Industry Relevance
Gene regulatory network inference is a critical step in target validation and mechanistic de-risking for biopharma discovery pipelines. The Inherent Dynamics Visualizer enhances predictive confidence by enabling interactive exploration of parameter impacts on network outputs, supporting data-driven hypothesis refinement. This capability improves translational continuity by aligning computational models with biological plausibility before preclinical investment.
Strategic Applications in Biopharma R&D
Early Discovery & Target Validation
- Scientific Value: Enables interrogation of therapeutic hypotheses through visualization of network topology changes under varying parameters.
- Operational Value: Supports biological de-risking by identifying parameter sets that yield biologically plausible network structures.
- Predictive Value: Increases confidence in target selection by revealing stable, oscillating network models that match experimental dynamics.
Screening & Assay Development
- Scientific Value: Provides standardized, visualization-ready gene lists and annotation files for downstream edge-finding workflows.
- Operational Value: Automates input file generation for pipeline steps, reducing manual configuration errors and improving reproducibility.
- Scalability: Enables rapid iteration across parameter sets to assess network robustness before compound screening campaigns.
Translational & Preclinical Research
- Translational Continuity: Facilitates alignment of inferred networks with known regulatory relationships from biological databases.
- Mechanistic De-risking: Allows removal of edges lacking experimental evidence to constrain network space to biologically feasible regions.
- Preclinical Alignment: Supports ODE modeling of networks using edge-derived parameters for in silico functional validation.
Pipeline & Workflow Integration
The tool integrates into the discovery continuum from early target hypothesis testing through lead identification by enabling parameter optimization at each inference step, improving model reliability before compound engagement.
- Discovery Biology: Supports hypothesis testing and pathway clarification via interactive visualization of node and edge identification outputs.
- Screening: Enhances assay readiness by generating standardized, reproducible input files for downstream modeling steps.
- Analytics: Enables quantitative comparison of network prevalence and stability across parameter conditions through edge prevalence tables and similarity scatterplots.
- Translational Research: Connects discovery-phase network models to preclinical validation by incorporating biological constraints from curated databases.
- Enterprise Reuse: Functions as a reusable configuration and visualization layer across multiple inference methods and time series datasets.
Operational & Enterprise Impact
- Scientific Value: Predictive confidence in target validation through visualization of parameter impact on network stability and dynamics.
- Operational Value: Standardization and reproducibility of configuration file generation and intermediate result evaluation.
- Strategic Value: Improved go/no-go decisions by reducing mechanistic ambiguity in early-stage network models.
- Portfolio Impact: Risk-adjusted prioritization of targets based on network robustness and alignment with known biology.
Implementation Considerations
- Requires expertise in gene regulatory network inference and time series transcriptomic data analysis.
- Depends on computational infrastructure to run the Inherent Dynamics Pipeline and host the web-based visualizer.
- Necessitates cross-team standardization of file naming conventions and parameter reporting for reproducibility.
- Involves adaptation considerations when applying the tool to different model organisms or data types beyond yeast cell cycle.
- Limited by the biological interpretability of DSGRN parameters, requiring integration with ODE modeling for functional validation.
