A general modeling and visualization tool for comparing different members of a group: application to studying tau-mediated regulation of microtubule dynamics

BackgroundInnumerable biological investigations require comparing collections of molecules, cells or organisms to one another with respect to one or more of their properties. Almost all of these comparisons are performed manually, which can be susceptible to inadvertent bias as well as miss subtle effects. The development and application of computer-assisted analytical and interpretive tools could help address these issues and thereby dramatically improve these investigations.ResultsWe have developed novel computer-assisted analytical and interpretive tools and applied them to recent studies examining the ability of 3-repeat and 4-repeat tau to regulate the dynamic behavior of microtubules in vitro. More specifically, we have developed an automated and objective method to define growth, shortening and attenuation events from real time videos of dynamic microtubules, and demonstrated its validity by comparing it to manually assessed data. Additionally, we have used the same data to develop a general strategy of building different models of interest, computing appropriate dissimilarity functions to compare them, and embedding them on a two-dimensional plot for visualization and easy comparison. Application of these methods to assess microtubule growth rates and growth rate distributions established the validity of the embedding procedure and revealed non-linearity in the relationship between the tau:tubulin molar ratio and growth rate distribution.ConclusionThis work addresses the need of the biological community for rigorously quantitative and generally applicable computational tools for comparative studies. The two-dimensional embedding method retains the inherent structure of the data, and yet markedly simplifies comparison between models and parameters of different samples. Most notably, even in cases where numerous parameters exist by which to compare the different samples, our embedding procedure provides a generally applicable computational strategy to detect subtle relationships between different molecules or conditions that might otherwise escape manual analyses.