Send to printer »

Tutorials : Jun 15, 2005 (Vol. 25, No. 12)

Taking Computational Biology to the Bench

Experiment Design Automation Informatics Software

Discovery is challenged by constantly changing data requirements and increasing analytics complexity. More and more labs need flexible informatics with less programming. New informatics software, called experiment design automation (XDA), can help labs improve the speed and quality of experimentation by providing design tools powered by data management and computational analytics to dynamically transform designs into lab automation without programming.

This ability for scientists to design and automate experiments with integrated analytics enables higher-value physical experiments that match the pace of research.

Current informatics platforms (i.e., databases or documents) are not able to meet XDA requirements. Databases provide scalability at the cost of expensive development, limited flexibility, and poor data mobility. Documents provide flexibility and mobility but are error-prone, labor-intensive, and not searchable.

XDA requires a third form of informatics based on models that provide the scalability and reliability of databases with the flexibility and mobility of documents. This model-driven informatics approach gives XDA the ability to meet the changing data and analytics requirements without programming.

Teranode (Seattle) Design Suite (TDS) enables labs to improve the speed and quality of experiments. TDS is a visual design tool that allows scientists to formally specify their work both in terms of the laboratory procedures and the biological systems of interest.

TDS captures a design as a computational model that includes data integration, data management, and analytics. As a computational model, experiment designs bring computational biology to the bench by enabling lab automation without programming.

TDS provides three applications that encompass the lifecycle of experiments from design though analysis. TDS Protocol Modeler enables design of lab procedures; TDS Protocol Player automates lab procedures; and TDS Biological Modeler provides pathway analytics.

Experiment Design

TDS Protocol Modeler is an application that allows a scientist to design an experiment electronically. Using a simple graphical interface, a scientist can sketch out a sequence of steps that define the procedures in the laboratory by dragging icons out onto the workspace and connecting the individual steps to form a sequence.

By graphically adding various properties to the schematic, the user can then define the measurements to be captured, the experimental parameters to vary, and the analytical processing to be performed.

Figure 1 shows the design of a simple compound screening assay, which includes steps that define the compounds to screen, the reformatting of the plates used to perform the assay, and the capture of expression data.

Data analysis steps can be integrated into the experiment protocol using TDS Protocol Modeler. Formulas defined at the plate or stack level can operate on data stored at the well level, greatly reducing errors and enabling real-time statistical processing, filtering, and analysis.

Here, an optimization routine has been used to analyze dose-response data and calculate EC50 values for a large set of molecules. Other protocol designs have included advanced statistical processing of data, processing of DNA and peptide sequence data, or image analysis and processing.

Using TDS Protocol Modeler, scientists can design experiment data and analytics requirements without programming.

Experiment Execution

One of the major benefits of formally specifying an experiment design in the TDS Protocol Modeler is that the experiment can be automated using a simple run-time environment called TDS Protocol Player.

TDS Protocol Player presents a wizard-guided interface that helps a laboratory technician execute the experiment in such a way that the experimental conditions are recorded; data is captured and semantically indexed; and audit trails are logged.

Figure 2 shows the Protocol Player interface to the screen assay previously discussed. The model specification includes the instructions to display to the lab technician at each step in the process, which data-entry fields are required, what reports are visible at runtime, and which branch of a conditional protocol should be taken based on real-time computational analysis.

Data entry and procedures can be a mix of manual and automated actions based on the design. In TDS Protocol Player, automation steps defined in the protocol design execute in sequence, but there is also the ability to define manual review points and to change the protocol to meet real-time changes.

Each time the assay is completed through the TDS Protocol Player, the data is stored using the semantics of the protocol design, allowing the results to be available for reporting and analysis using the full context of the experiment parameters, samples, and procedures.

Pathway Analysis

Ultimately, a scientific experiment is performed because a scientist expects that experiment to provide some insight into the functioning of a real biological system. Achieving deep understanding requires experiment data from many different methodologies to be integrated and evaluated in the context of a biological pathway.

This form of pathway analytics has traditionally required programming in math languages and manual curation of data, which has made computational biology nearly impossible for most bench scientists.

TDS Biological Modeler is a visual application to import biological models and schematically specify biological pathways for analyzing experimental data.

TDS Biological Modeler uses the same underlying model-driven informatics platform as TDS Protocol Modeler to enable the direct linkage between experiment data collected in the lab and analysis of data in biological pathway models.

TDS Biological Modeler is designed to allow easy reuse of pathway models and data from many sources, including KEGG, SBML, Affymetrix, and MAGEML.

Figure 3 shows a model of the EGF signaling pathway in TDS Biological Modeler. Rate constants, references, and expression data can all be stored in the same model of the pathway, and easily exported into other analysis packages.

TDS Biological Modeler's chemical network simulation capabilities have also proved useful for modeling and refining experimental assays. By providing detailed kinetic simulations of expected assay behavior, an assay model can validate that an assay is performing as expected, and optimize assay design by simulating the response to different experimental conditions.

To specify a series of chemical reactions in Biological Modeler, a scientist needs only to drag and drop icons onto the canvas to represent the chemical species and reactions of interest, define the connections, and specify the rate constants and mechanism of each reaction.

TDS Biological Modeler can then automatically translate the graphical model into an ODE-based mathematical representation and perform a simulation of the system's kinetics. This feature makes kinetic modeling more accessible to the large fraction of bench scientists lacking a formal training in mathematical modeling.

Additionally, the package provides several advanced model-based analysis capabilities. For example, a scientist might explore a model's behavior by scripting a large number of different simulation runs, and then saving the parameters for the most interesting simulations for more detailed study.

Built-in optimization algorithms allow scientists to estimate parameters of assay or biological models using actual experimental data. Finally, the TDS allows computational biologists to add their own analysis algorithms to the TDS Biological Modeler.

Discovery is challenged by constantly changing data requirements and increasing analytics complexity. Teranode Design Suite provides XDA functionality to enable labs to flexibly design and automate experiments. TDS accomplishes this by providing data management and computational analytics in a visual environment without the need for programming.

The benefit is that scientists can use computational biology at the bench to enable higher-value physical experiments that match the pace of research.