(517j) An Ontological Framework for Knowledge Modeling in Pharmaceutical Product Development
Development of a drug product is a complex, iterative process consisting of selection of a dosage form, excipients, processing route, operating equipment and so on. At each stage, knowledge in various forms, including heuristics, decision trees and mathematical models, is used in making decisions. Typically knowledge is modeled specifically for the tool that uses it, such as expert systems and simulation tools. This makes it very difficult to share the knowledge across different tools and among development teams, and integrate various forms of knowledge to assist in making pharmaceutical product development decisions. To provide easier access to available knowledge and better decision support, we propose an integrated approach to systematically model the different forms of knowledge.
An ontology based approach is used to model knowledge in the form of guidelines. A guideline models procedural knowledge, which consists of decision logic, information look-up, evaluation of decision variables and making recommendations. For example, to determine whether direct compression is appropriate for a particular drug substance, values of several properties, such as flowability, compressibility are examined. A guideline contains five basic steps: action, decision, state, branch and synchronization. For example, the decision step Is_Flowability_Very_Bad is based on expression Hausner_Ratio > 1.5 which has details to access the Hausner ratio of desired material. Guidelines are developed based on the knowledge captured from detailed discussions with faculty in Department of Industrial and Physical Pharmacy at Purdue University and experts from various pharmaceutical companies. In order to provide the decision support, a java based engine is developed for guideline execution, which uses knowledge in the guideline and information from a repository based on the domain ontology.
Reformulation of a MDRTB (multi-drug resistance tuberculosis) drug is used as a case study to demonstrate the applicability and benefits of the proposed approach. The guidelines are used for recommending processing route and route dependent excipients to manufacture the drug product as immediate release oral solid dosage form. To select excipients these guidelines use mixing rules to predict the mixture properties such as flow properties. The mixing rule that computes the mixture properties as the weighted average of pure component properties are used. Mixing rules and other mathematical knowledge are modeled in form of ontologies and accessed directly by the guideline execution engine.
This approach provides an open and easy way to create, use and modify the knowledge and supports the integration of the various forms of knowledge and information. A simple but complete representation model for guidelines makes it easier for the domain experts to interact directly with the guidelines and add/modify them based on the requirements. The proposed approach could also be used to effectively capture the knowledge in different domains including process development and operations in chemical and pharmaceutical industry.
Keywords: Pharmaceutical Product Development, Formulation, Knowledge Modeling, Decision Support, Guidelines, Ontology