Epithelial tissue statistics: eliminating bias reveals morphological and morphogenetic features.

Geometric order in quasi-two-dimensional epithelia has been extensively researched in order to identify and classify different tissues to help our understanding of how tissues form (morphogenesis) and how their formation may be influenced (tissue regeneration). However, the significance of published data -such as the distribution of numbers of cell neighbors- has been debatable because of measurement bias. We shown that such bias can be detected and corrected without detailed knowledge of the original samples, using only the biased (measured) distributions. This is true for both of the most important sources of bias: the measurement of apparent four-fold vertices and the selective preference for measuring smaller cells introduced by selecting a finite sampling window. The resulting unbiased data allows for a meaningful comparison of all available data, from different sources, taken with different experimental resolution and methodology. Conclusive evidence is found that the apparent four-fold vertices are neither distributed randomly nor oriented randomly, revealing profound differences in topological correlation between proliferating and remodeling tissues. The method is applied to measurements of Drosophila wing tissue, where it successfully disentangles distributional moments, allowing for an assessment of their relative importance, independence, and significance in tissue identification and classification.