Can Better Modeling Reduce Pharmaceutical Development and Manufacturing Costs?

In the UK, a new four-year initiative has been launched to streamline drug development and manufacturing by leveraging better computer-based process modeling and simulation. Dubbed Advanced Digital Design of Pharmaceutical Therapeutics (ADDoPT), the $29-million program, part of the UK's Advanced Manufacturing Supply Chain Initiative, aims to develop the tools required for an “efficient, knowledge-based, quality by design-oriented [pharmaceutical] supply chain.”  Sponsoring the effort is the Medicines Manufacturing Industry Partnership an alliance between The Association of British Pharmaceutical Industry and the UK BioIndustry Association.

Astra-Zeneca, BMS, GSK and Pfizer are participating in the program, as are the universities of Cambridge, Leeds and Strathclyde, particularly its Center for Innovative Manufacturing in Continuous Manufacturing and Crystallization.

Also collaborating are the Cambridge Crystallographic Data Center, said to be the world’s largest repository of crystallization data, and the Hartree Center, a center for high-performance computing operated by the UK’s Science and Technology Facilities Council.  

Participating vendors include Process Systems Enterprise (PSE) Ltd., a specialist in process modeling and simulation, the process-control company Perceptive Engineering, and Britest, a nonprofit organization that focuses on improving process understanding and value through better use of modeling.

The project’s loftier goals are to develop new approaches that would allow pharmaceutical manufacturers to deliver medicines more effectively to patients, to use data analysis and first principle models to better define, design and control pharmaceutical manufacturing processes.

On a more pragmatic level, ADDoPT also aims to develop a strong knowledge-based workforce and centers of excellence within the UK, where the impact of recent downsizings have had ripple effects on the economy, most notably the shuttering of Pfizer’s 60-year-old R&D facility in Sandwich, Kent, which had employed more than 2,000 scientists and technicians. 

As of 2013, there were 380 pharmaceutical companies in the UK generating more than $43 billion in revenues.  According to the program’s sponsors, ADDoPT “has the potential to propel the UK to the forefront of medicinal product design and manufacturing.”

One of ADDoPT’s goals will be to diversify the modeling techniques available to pharmaceutical manufacturers, says Sean Bermingham, principal consultant and vice president of PSE’s solids division, who is the project’s technical lead.  PSE was established as a spinoff of Imperial College UK, in 1997 to develop process simulation and modeling software for the chemicals and oil and gas sectors.  In 2010, it set up a pharmaceutical and life sciences division to bring this capability to drug companies.  A growing number of Big Pharma companies are using the software, and the division currently accounts for one quarter of PSE's business, Bermingham says.  One focus is on applying modeling to support continuous processing initiatives. Use of mechanistic models can be particularly important in this area.

Currently, many pharma companies rely on statistical modeling and design of experiments (DoE) approaches in their drug development programs. However, DoE can often require a large number of experiments, and considerable resources.  Mechanistic modeling would find correlations between different variables, so that fewer, more targeted experiments could be run. 

For instance, Bermingham explains, instead of doing 20 different spray drying experiments, one would be able to start with vapor sorption, estimate isotherm parameters, then work on kinetics using single droplet studies, using measurements to describe residence time, then move on to spray drying with only two or three experiments.  “It requires a lot less material and results in data with higher confidence limits than DoE,” he says.

However, using this approach will require a very different mindset and new workflows than many manufacturers currently use.  Tackling those challenges, and the training involved, is an important aspect of ADDoPT, he says.  The project will also use Big Data techniques and  high-performance computing available via the Hartree Center, to develop equations in areas where science hasn’t advanced enough to allow the use of mechanistic modeling, Bermingham says.  An example would be material properties. 

“You may run tests on a compactor simulator, but when you change the excipient, blend or the particle size distribution within the API, the result will change. Our research would aim to generate a compressability equation,” says Bermingham.  This predictive equation would describe compressability regardless of specific conditions, allowing manufacturers to model more effectively.

Bermingham says that Big Pharma’s participation will be critical, and will provide real-world measurements, for example, of different compressability figures in different blends, that would be used to develop mechanistic models.   “We hope that, by applying Big Data techniques, we will be able to find overarching correlations, and predictive models,” he says. 

Another issue that the project will study closely is crystallization and polymorph formation, using the Cambridge Center’s database as well as powerful data mining techniques. The result could be models that would allow developers to be able to predict polymorphs for new drugs.

The efforts could have considerable impact, not only on R&D but on scaleup and tech transfer, allowing companies to weed out drug candidates with less scaleup potential, much sooner, and to focus on those that can be scaled up successfully.

Source