An ontological informatics framework for pharmaceutical product development: Milling as a case study
Pharmaceutical product development is an expensive, time consuming and information intensive process. Providing the right information at the right time is of great importance in pharmaceutical industry. To achieve this, knowledge management is the approach to deal with the humongous quantity of information. Ontological approach proposed in Venkat et al.  to achieve the knowledge management in pharmaceutical product development. The abstract divisions of 'Data', 'Information' and 'Knowledge' are used to create a decision support for pharmaceutical processes. In particular, pharmaceutical milling process knowledge has been modeled using ontological framework. Along with knowledge modeling, mathematical modeling of milling process is performed. The current work has two aspects: 'Knowledge modeling of milling process' and 'Model development and experimental validation of the milling process'. Ontology for milling process and guidelines for milling equipment selection have been handled in the current work. Data driven mathematical knowledge is modeled using the ontological framework.^ As a case study for model development using ontological framework, a hybrid mill model is developed. The hybrid neural network - population balance model predicts the particle size in milling. Experimental work has been done on a lab scale Quadro comil to verify the hybrid model. Ribbon properties such as ribbon density and tensile strength are measured. The milling experiments are performed on three types of API (Active Pharmaceutical Ingredient) ribbons altering the operating conditions. The hybrid model, where neural networks function as the breakage kernels are combined with the traditional one dimensional discrete form of population balance model. The material properties and operating conditions become the input nodes for the network and the outputs being the breakage functions.^ The hybrid model is a case study for the model development aspect of decision support system. The experimental data is stored as instances of experimental ontology. Using the mathematical (neural network, population balance etc.), material and experimental ontologies the model is built in the decision support. Not only the mathematical model building, but also the heuristic knowledge is also modeled using the technology of 'Guideline Modeling' called GLIF (GuideLine Interchange Format). Mill equipment selection is shown as a case study for modeling heuristic knowledge. The decision support built is a generic framework to accommodate any chemical and pharmaceutical unit operation, in general. ^
Venkat Venkatasubramanian, Purdue University, Gintaras V. Reklaitis, Purdue University.
Chemistry, Pharmaceutical|Engineering, Chemical|Information Science
Off-Campus Purdue Users:
To access this dissertation, please log in to our