Figure 4. Biomedical terminologies are likely missing the vast majority of domain-specific, synonymous relationships.

The numbers of undocumented concepts and synonyms specific to each biomedical sub-domain were estimated using a hierarchical mixture model in order to capture annotation variability that occurred within and across terminologies (10 concept components, each with 4 synonym components, see Materials and Methods and Supporting Information Text S1). In panels (A) and (B), the number of documented concepts per component (green, above x-axis) is compared to the estimated number of undocumented concepts per component (blue, below x-axis): (A) Diseases and Syndromes and (B) Pharmacological Substances. In panels (C) and (D), the number of documented synonyms per mixture component (green, above x-axis) is compared to the estimated number of undocumented synonyms, which come in two flavors, undocumented synonyms paired to documented concepts (blue, below x-axis) and undocumented synonyms paired to undocumented concepts (red, below x-axis).