Probabilistic Knowledge Assembly (PANDA) Ontology has been developed in the context of the DARPA's Big Mechanism research program. It provides a semantic interface for capturing information on molecular interactions extracted from scientific texts. Conceptually, PAF ontology supports 3 layers of representation: 1) Domain knowledge: molecular interactions, their participants, structure, and (possibly) formal relationships to some external biological models. 2) Textual evidence: sentences extracted from text, speaking about those interactions, and their provenance-related information. 3) Probabilisitic information associated with different aspects of the representation (see the imported Uncertainty Ontology).