The role of solubility modelling and prediction grows ever more crucial in today’s drug development process, argues Gary O’Neill, director of business consulting, Pharmaceutical Business Unit, AspenTech
During the pharmaceutical drug development process, manufacturers must create a solvent system in which the reactants are sufficiently soluble to drive the chemical reaction that creates the end product. Following that creation, the product initially remains dissolved in the solvent system.
The next stage of the process, known as solvent swaps, involves the existing solvent system being replaced, usually through distillation. New solvent is added gradually to the mix, creating conditions in which the desired product is insoluble while all other reactants and by-products remain soluble.
Solubility is key to the drug production process both at the initial solubilisation of reactants and at the back end in bringing the final pharmaceutical product out of solution in as pure a form as possible. The front end process is straightforward because the manufacturer will typically be familiar with all the reactants and their properties. The back end, however, is far more complex as little is likely to be known about the properties of the new chemical entity (nce), and even less about its likely solubility.
This presents a problem. Manufacturers need to find out how soluble an nce is in potentially hundreds of different solvents or solvent systems to ensure the purity and manufacturability of a given nce. Unfortunately, they typically lack the time or resource to carry out all the necessary measurements, even with high throughput experimentation.
Some may end up continuing the development process without having full confidence in the end result. With only small quantities of nce typically available in the late stages of drug discovery, there is little opportunity for determining solubility in the lab, so characterisation typically occurs later in the development cycle.
Unfortunately, by that time the laboratory production method for the candidate drug may turn out to be difficult to scale up while maintaining acceptable yield and required purity.
finding an answer
Today, solubility modelling and prediction techniques have the potential to resolve these dilemmas and increasingly have a critical role to play in ensuring the quality and manufacturability of new pharmaceutical products.
In their most effective form, these techniques allow drug researchers and process development engineers to take small amounts of the nce and test solubility in as few as four pure solvents. In turn, these results enable the researchers to predict solubility in hundreds of other solvents and solvent systems – without having to carry out experiments in each system. This drives faster time to market and minimises production time and costs.
The approach helps firms working on new drugs to select the most appropriate solvent system for a given compound. It can help decide which candidate nces are taken forward into formal development and how to process a product in manufacturing, enabling drug manufacturers to evaluate manufacturing risk and improve confidence during the development process, while at the same time resulting in cost and productivity benefits.
The Aspen Solubility Modeler product, for example, helps pharma manufacturers save time and money by ‘projecting’ solubility in a range of other solvents and solvent systems based typically on real measurements of pure-component solubility in four or five solvents.
Of course, no drug company can use those numbers as an absolute guarantee, but where solubility modelling and prediction really bring value is by eliminating the hundred or more different approaches that are unlikely to work, and helping to identify the two or three that are most likely to be successful.
downstream benefits
Solubility modelling and prediction becomes increasingly important as the drug discovery phase merges into process development. At this point, the number of nces is reduced to a handful – in some cases with each offering roughly the same therapeutic value. This process can help manufacturers decide from a business standpoint which of these drugs will be easiest to produce.
The approach, enabled by AspenTech’s recently patented non-random two-liquid segment activity coefficient (NRTL-SAC) model, has a range of benefits.
One leading manufacturer recently found that it was able to double its throughput of nces evaluated in process development without adding any people. In other words by cutting back significantly on experimentation time, the same number of staff was able to do twice the work that had previously been possible.
The approach also allows manufacturers to begin process development activities and, often, production for clinical trials sooner using the predictive modelling capabilities of the software to quickly design optimised manufacturing processes.
Another key benefit is the ability to reduce time and effort when drug production is transferred to the pilot plant, and ultimately to commercial production. The approach enables informed design of the final Active Pharmaceutical Ingredient (API) manufacturing processes so they can be designed to use solvents already in use at the target manufacturing site.
In line with Quality by Design (QbD) concepts, the process also helps manufacturers to comply with the latest regulatory initiatives. Regulators are focused on driving up levels of understanding of both drug properties and manufacturing practices. Decisions have to be based on a sound understanding of drug and process. Manufacturers need to ask themselves – what are the critical quality attributes that are required for a drug to be effective and what manufacturing parameters are important to achieve those critical quality attributes?
Solubility modelling gives manufacturers a scientific basis for the decisions that they make so that as a drug makes its way through the regulatory approval process, they not only have data around their solvent system but also a history of why they chose that system. This builds a much stronger technical case to present to the regulatory authority during the approval process.
Solubility Modeler is part of AspenTech’s aspenONE for Pharmaceuticals suite and pharmaceutical companies are already using it to reduce r&d time and costs and to ensure a highly repeatable manufacturing process. For example, using the software enables GSK to understand thoroughly the solubility properties of nces to assess risks more quickly and improve confidence during the development of new pharmaceuticals.
essential component
‘Screening for crystalline forms is an essential component in pharmaceutical drug development,’ says Stephen Carino, investigator, Solid Form Sciences Group, GlaxoSmithKline. ‘In our high-throughput screening workflow, we have utilised NRTL-SAC in Aspen Properties to predict the equilibrium solubility of the drug in single- and multi-component solvent systems. The predicted solubility values are used in selecting an appropriate set of solvent systems that are explicitly unique for each of the crystallisation modes. This rational solvent selection coupled with the systematic screening approach has allowed us to assess risk around solid forms and improve our confidence in the robustness of the API processes.’
According to Peter Crafts, principal process engineer, AstraZeneca Pharmaceuticals, solubility prediction is of strategic importance to AstraZeneca Process R&D: ‘Our industry is increasingly challenged to deliver efficient and environmentally acceptable processes. Taking just a few physical property measurements and extrapolating them to predict temperature and composition effects on solubility is very desirable.
‘The solubility modelling capability with NRTL-SAC in aspenONE V7 is one of the tools we use in our crystallisation process design workflow. aspenONE helps leverage solubility data to optimise solvent choice and processing conditions. We look forward to future developments from AspenTech that will further increase the areas of application.’
Looking to the future, the next evolution of solubility modelling is likely to be an expansion of the thermodynamic database of solvents available for solubility prediction.
That expansion is likely to include electrolyte solutions. Small molecule NCEs are typically produced through a series of organic chemical reactions most often taking place in organic solvents. However, electrolyte systems offer the possibility of using insoluble salt forms for crystallisation of either the target drug or undesired by-products. The addition of electrolyte solvent systems to the NRTL-SAC database could then further expand the capabilities of the Solubility Modeler tool.
It also means that, in the future, products that rely on electrolyte systems in the manufacturing process may also be able to benefit from the predictive power of solubility modelling.
