Key data to model the sustainability of the drug production process

The environmental impact of drug production has been in the spotlight in recent years. Initially, the small molecules sector was at the center of reports linking poor wastewater treatment at factories making antimicrobial APIs to the spread of antibiotic resistance.

While the environment concerns remain, the focus has recently expanded to broader ideas of sustainability and economics. Much of the attention has been focused on process development, as scientists strive to understand the impact of ingredients, raw materials and energy requirements on costs and waste generation.

For small molecule production processes with few inputs and outputs, modeling sustainability is straightforward, according to Oliver Fisher, PhD, a researcher at the University of Nottingham in the UK. However, for companies manufacturing complex biopharmaceuticals, the challenge is much greater.

“The results of a biopharmaceutical process can include products, co-products, by-products and waste streams,” says Fisher. “Each result can have several, often contradictory, sustainability implications.

“Therefore, a model whose sole objective is to maximize a single outcome, for example product yield, cannot assess the implications on the overall economic or environmental performance of the process without understanding what effect maximizing product yield has on the rest of the process results. “

Neural network

In an attempt to resolve this problem, Fisher and his colleagues used a neural network—Bond nodes capable of receiving and analyzing multiple data inputs — to generate a predictive sustainability model that they claim to be more accurate than established approaches.

“Traditional approaches such as the first major modeling require a high level of understanding of the underlying physics of how systems work. Data-driven models, however, are derived from fitting process data to algorithms such as neural networks and require less knowledge of the detailed mechanisms underlying the process, ”he explains.

“Data-driven models can actually take advantage of the increasing volume of data being produced, combined with increasing computing power. There is a huge opportunity to extend the reach of modeling across the process and supply chain to better assess process sustainability.

Neural networks have already been used for the assessment of the sustainability of processes. However, the approach developed by Fisher and his colleagues uses a network of networks which, he says, allows for a more precise model.

“Rather than using a single neural network to simultaneously predict all outputs, which suffered from data overfitting, we explored the development of a chain of models using the whole chain method. regressors, ”he said. GEN. “Overfitting is a real challenge in predictive modeling, [While things are fine where the fit is great for a limited set of data], but when more data is added to the model, it either fails or is unable to reliably predict future scenarios.

“A regressor chain builds a series of models where each model is built using the output of the previous model as input for the next. The set of regressor chains works by creating multiple regressor chains for each permutation of the output sequence order. This method captured the relationships between process outputs, providing an understanding of the ripple effects of changing one output on others.

Monitoring requirements

High-quality data is essential, according to Fisher, who says biopharmaceutical companies looking to use a neural network to predict sustainability will need to ensure they can monitor manufacturing processes in detail.

“As with all data-driven models, process limits and variability should be defined in the input and output data from which multi-target models are derived, because waste equals waste,” he continues. “You have to look at the data available and [determine] whether the volume and granularity of the data is sufficient to describe the process.

Pharmaceutical companies may need to install additional in-process monitoring systems, emphasizes Fisher, who notes that “a multi-target model may require more measured output data to label the input data, which is may prove to be more expensive initially if this output data is not collected originally. However, the model used to improve the economic and environmental sustainability of the process would prove useful in the long run.