Life Sciences Data Analysis

Abstract

The Life Science Workbench is a software suite from MDL Information Systems that helps biologists create and track protocols, capture experimental results, analyze data, integrate results, and provide extensive functions for screening data. The Data Analysis Toolbox (DAT), reviewed here, is an add-in to Microsoft Excel that performs the curve-fitting and data analysis functions of the suite. The program contains more than 40 models of physical processes including common linear, nonlinear, logarithmic, power, and polynomial families of curves. The advantage of add-in programs such as this one is convenience, since the functions of Excel are known to most users. DAT was installed within seconds on a 166-MHz Pentium and integrated seamlessly with Excel (requires version 7.0 or newer for Office 95). The software places an “Analyze” feature, containing six functions, within the “Data” group on the tool bar. Users may select data for an analysis set by defining criteria (filters), such as data position, numerical size of the data set, lack-of-fit, and data distribution. The position and spread of the data can be filtered by defining a set point, upper and lower bounds, or difference from an estimated response. DAT reacts when these limits are exceeded, but unfortunately, not in a very useful fashion. No warning box appears on the screen. The calculations proceed unhindered, and the analyst must search for the output tab that includes a long row of ones and zeros to view disqualified data. The immediate appearance of a warning box stating the problem would be helpful. A Wizard assists in graph construction. In the first step, a data group is chosen from categories that include enzyme kinetics, ligand-receptor binding, exponential, linear, logarithmic, and polynomial. Users next choose a specific model within a data group. For example, models available from the Michaelis-Menten subgroup include steady-state, random two-reactant system, or substrate plus inhibitor. One can display the equation of the final, selected model and set a weighting option for the data. Subsequent steps define set-up parameters for input data and data selection, which, as in other Excel routines, may involve clicking cells or typing into a dialog box. Step four permits selection of the desired statistics: parameter estimates, standard error, confidence values, and analysis of variance tables. Finally, graph type is specified. An experienced user can complete the process in less than a minute and use Excel's functions to format output. After the graphics are completed, a “Back Calculate” feature generates revised output values for all input values, even those out of the original range. The “Preferences” box sets the value for convergence, number format, and confidence level. A registration booklet and a user's guide are provided with DAT. Although the latter contains sections that are not completely clear, the examples are useful. The only noteworthy drawbacks to DAT are an unwieldy visual statistical output (stretched out over many columns rather than in a compact table) and the lack of P values for the F-test. The P value is the statistician's prized comparative metric, so this omission will not only be a problem to the novice user, who may not know how to calculate it, but a nuisance to the more experienced analyst, who comes to expect it.