Recently Bowers et al. [1] analyzed triplet logic relationships among 4873 Clusters of Orthologous Groups (COGS) from 67 fully sequenced organisms by calculating how well logic relationships between proteins a and b predicted the presence or absence of protein c (the uncertainty). The log of the normalized uncertainty distribution follows an approximately linear relationship for uncertainties in the interval [0.1, 0.9]. Using fitted parameters of this relationship as a characterization, we develop four types of visual analysis for LAPP data: distributions of uncertainty over logical relation type, distributions of uncertainty over functional categories, relationships of uncertainty of the overall population to known network relationships of a particular organism, and relationships of uncertainty distributions to groups obtained by standard clustering techniques. The purpose of this study is two-fold: to better understand the implications of uncertainty predictions for automatic protei...
Kay A. Robbins, Li Zhao