1Department of Paediatrics, Division of Infectious and Immunological Diseases, Child and Family Research Institute, University of British Columbia, 2Department of Computer Science, University of British Columbia, 3Department of Psychology, University of British Columbia
This article is a part ofJoVE Immunology and Infection. If you think this article would be useful for your research, please recommend JoVE to your institution's librarian.Recommend JoVE to Your Librarian
Current Access Through Your IP Address
Current Access Through Your Registered Email Address
Shih, D. C., Ho, K. C., Melnick, K. M., Rensink, R. A., Kollmann, T. R., Fortuno III, E. S. Facilitating the Analysis of Immunological Data with Visual Analytic Techniques. J. Vis. Exp. (47), e2397, doi:10.3791/2397 (2011).
Visual analytics (VA) has emerged as a new way to analyze large dataset through interactive visual display. We demonstrated the utility and the flexibility of a VA approach in the analysis of biological datasets. Examples of these datasets in immunology include flow cytometry, Luminex data, and genotyping (e.g., single nucleotide polymorphism) data. Contrary to the traditional information visualization approach, VA restores the analysis power in the hands of analyst by allowing the analyst to engage in real-time data exploration process. We selected the VA software called Tableau after evaluating several VA tools. Two types of analysis tasks analysis within and between datasets were demonstrated in the video presentation using an approach called paired analysis. Paired analysis, as defined in VA, is an analysis approach in which a VA tool expert works side-by-side with a domain expert during the analysis. The domain expert is the one who understands the significance of the data, and asks the questions that the collected data might address. The tool expert then creates visualizations to help find patterns in the data that might answer these questions. The short lag-time between the hypothesis generation and the rapid visual display of the data is the main advantage of a VA approach.
1. Exploration-based Analysis on Tableau
2. Presentation-based Needs
3. REPRESENTATIVE RESULTS
Figure 1. A screenshot of Tableau after importing the spreadsheet named NFKBIA from the Excel file demo.xls. The dimensions and measures shelves were properly populated with the categorical and numerical data, respectively.
Figure 2. The Calculated Field window is invoked to create a special calculated field to use in Tableau. The list on the bottom left-hand box helps identify possible fields, and the list on the right-hand side contains abbreviation of functions that can be used in the formula. In this example, we wanted to add the values for PFD4, PFD3 and PFD2 to obtain the final value that we refer to as PFD > 2
Figure 3. Visualization of stimulus concentration level vs. observed cytokine concentration. The visualization shows a plot of the different concentration levels of the stimulus 3M-002 against the observed concentration of the cytokine TNF-α. The colors of the lines refer to the different genotypes for a single-nucleotide polymorphism in the NFKBIA gene of the individuals in our innate immune study.
Figure 4. A screenshot of a two-column visualization matrix. We generated a two-column matrix to facilitate a side-by-side comparison of responses to two stimuli, 3M-003 and LPS. The x-axes are the different concentration levels of the two stimuli, and the y-axis plots the values of the calculated field, PFD > 2.
Figure 5. These Tableau dialogue windows illustrate how to connect data recorded in different spreadsheets. Connecting data from different spreadsheets can be accomplished by combining these using logical join clauses of key values.
|Visualization and Analysis Tool|
|Parallel coordinate plots||Yes||Yes||Yes||Yes||Yes||No||Yes||Yes||No||No|
|Scatter plot matrixes||Yes||No||Yes||Yes||Yes||No||Yes||No||No||No|
|Direct manipulation of data||Yes||Yes||Yes||Yes||Yes||Yes||Yes||No||Yes||No|
|Extensibility to other platforms (e.g., R)||Yes||No||Yes||Yes||No||Yes||No||Yes||No||No|
|CSV table formats||Yes||Yes||Yes||Yes||No||No||Yes||Yes||Yes||No|
|XML data formats||Yes||No||No||Yes||Yes||No||Yes||Yes||Yes||No|
|Can deal with 10000+ rows||Yes||No||No||No||Yes||No||No||No||No||Yes|
Table 1. List of visual analytics tools and some of their features.
The advent of high-throughput technology in modern biomedical research led to an explosion of research data that requires a more efficient way of analysis. Visual analytics (VA) is the science of analytical reasoning facilitated by interactive visual interfaces (1). The VA approach restores the analytical power in the hands of human analyst, contrary to the traditional approach to detect patterns by computer. Visual analytics has been applied to research in various fields, such as defense research (1) and hurricane trends (2). So far, there are only a few examples of VA applications in biology (3). We demonstrated in this video article that VA is an approach that can be added to the biologist's arsenal of analysis tools. Many VA softwares are available ranging from those that are in development in academic labs to those that are commercially-available. For our work on neonatal innate immunity (4), we chose Tableau because of its suitability in analyzing spreadsheet-style datasets available in the lab. Other VA tools, some of which we mentioned in our video article, may be more appropriate for other types of biological data. We listed the functions and characteristics of some of the more popular VA tools in Table 1. This list is not meant to be exhaustive because it is beyond the scope of our study, but it should be a good starting point for scientists to determine the suitable VA tool for their specific datasets.
There are two major points about VA that we would like to highlight. One, the VA approach is intended as an exploration process by helping the analyst quickly spot patterns such as general trends and outliers in the data. The main focus of VA is to provide a powerful visualization technique for large datasets. It is not an alternative to statistical analysis. In fact, most of the VA tools are very limited in their ability to perform statistical analysis although we anticipate this to change in the near future. The second point we want to mention is that the data pre-processing prior to importing the dataset to a VA tool is crucial for the success of the analysis. Bear in mind that data presented in a human-readable fashion in spreadsheets are sometimes different from a machine-readable format. Data pre-processing can be a time-consuming step, often requiring massive transformation of the data, if the original format of the data is not suitable to the VA tool. We highly recommend the careful planning of data entry and encoding to allow flexible and efficient uploads to downstream analysis software tools such as those available in VA.
No conflicts of interest declared.
We would like to thank the members of Vancouver Institute of Visual Analytics (VIVA) for offering comments and advice for the project. In particular, we would like to thank John Dill, Brian Fisher, and David Darvill. We would also like to thank the members of the Kollmann lab for their support and helpful discussions. This work was supported in part by National Institute of Allergy and Infectious Diseases, National Institute of Health Grant N01 AI50023; AllerGen NCE Grants 07-A1A and 07-B2B; and the Michael Smith Foundation for Health Research. T.R.K. is supported in part by a Career Award in the Biomedical Science from the Burroughs Wellcome Fund ad by a Canadian Institutes for Health Research Training Grant in Canadian Child Health Clinician Scientist Program, in partnership with Sick-Kids Foundation, Child and Family Research Institute (British Columbia), Women and Children's Health Research Institute (Alberta), and Manitoba Institute of Child Health.