Conventional ontologies comprise deterministically organized concepts. Certain ontological relations (e.g., those between diseases and causes, symptoms, treatments, or prognosis) cannot be represented faithfully without handling uncertainty. Biomedical ontologies are useful, only if they point to outcomes of the phenomena of interest. Such outcomes are usually associated with probabilistic data. This study is built upon Bayesian probability theory and machine learning where determinism is treated as a special case over a set of probabilistic ontological relations.