How computer simulations can help predict microfluidics experimental outcomes

Microfluidic technologies are rapidly becoming the method of choice for micro- and nanoparticle production across a number of sectors, from cell biology research to drug development. This approach offers several advantages over traditional batch methods, including precise process control, enhanced reliability, and the ability to easily scale up production. This article discusses how computational fluid dynamics modelling and artificial intelligence approaches are enabling scientists to predict the outcome of microfluidic experiments before putting new protocols into practice, reducing the time taken to set up and optimise novel experiments.

Microfluidic nanoparticle synthesis 

Synthetic polymeric nanoparticles are being increasingly used for drug delivery applications in several fields of nanomedicine as well as in clinical and basic research. Poly(lactic-co-glycolic acid) – PLGA – is one of the most successfully developed nanoparticles to date due to its safety, biocompatibility and biodegradability, and has been approved by the US Food and Drug Administration (FDA) for use in several medical applications, including drug encapsulation. The generation of these particles, however, requires incredible precision; for example, a well-controlled bead size distribution is essential for drug delivery vehicles, as particle size and a lack of monodispersity significantly influence drug release rates. This is where microfluidics is extremely useful, offering exceptional reproducibility and tunability of particle generation, with the advantage of reduced reagent consumption and less waste compared to more traditional batch synthesis. 

Microfluidics has been successfully used to encapsulate a number of drugs, including paclitaxel and doxorubicin, into PLGA nanoparticles. Microfluidic devices, or chips, enable the production of droplets, particles or emulsions of consistent sizes through adjustment of physicochemical parameters such as flow rate, temperature, and reagent composition and concentration. However, this can be time-consuming, typically requiring an extensive period of laboratory optimisation. This is where in silico technologies can bring major benefits, allowing researchers to predict the outcome of a particle manufacturing process according to fluid properties, such as viscosity and surface tension, prior to running the experiments in a laboratory. Microfluidics and computer scientists are now starting to build productive partnerships with the aim of developing computerised tools that can simulate and predict particle behaviour according to a specified set of variables, ultimately saving on time, reagents and manual labor. 

The power of cross-functional collaborations

Scientists are no longer staying isolated in their niche research areas but are increasingly joining forces with experts in other fields to tackle complex physics and engineering problems. In the field of nanoparticle synthesis, computerised tools can aid in predicting the specific size, morphology and other physicochemical properties of polymeric nanoparticles through the use of complex mathematical equations and models. Two approaches that are now being developed to tackle these issues include computational fluid dynamics (CFD) and artificial intelligence (AI). CFD uses numerical analysis and algorithms to simulate and predict fluid flow behaviour, whereas AI functions independently and intelligently through a machine learning approach. Both aim to help bench scientists predict the outcome of particle generation protocols, without the need to run the experiments in the lab. 

CFD for accurate prediction of microfluidic bead production

CFD modelling works by processing known physical and mathematical equations for solving complex fluidic problems. Based on this premise, a unique software – MOEBIUS (Lexma Technology) – was developed that could simulate the production of PLGA droplets in Aqua-Phase and accurately predict trends in bead properties. This software was tested in a real-world experiment and proven to accurately replicate existing manufacturer’s data. Using this software, scientists can now test a variety of combinatorial pathways by changing variables such as fluid concentration, viscosity, surface tension and flow rates, as well as channel geometries and coatings, to infer an optimal protocol to use as a starting point in the lab, minimising the need for additional optimisation. Computer simulations such as this one will be of real use in future for supporting research labs and the biotech industry.

AI applications for rapid manufacturing of size-tunable PLGA microparticles 

Another approach that scientists have evaluated for predicting the sizes of monodisperse PLGA microparticles generated by microfluidics is AI, more specifically a subset of AI called machine learning. Unlike CFD simulations, this model uses a supervised learning approach to train an artificial neuron network (ANN) inspired by biological neurons to interpret the non-linear relationships between various experimental parameters – flow rates, polymer concentrations, etc. – and their effect on particle generation. This approach is generally preferred when the physics of the processes involved are not well-known. Instead of providing the computer with a mathematical formula, it is fed with a large amount of data regarding different experimental parameters and outcomes, to enable it to make predictions based on its learnings from those datasets. Once tested and validated, the model can be used to process external datasets containing previously unseen information, to assess its capability to predict experimental outcomes. Researchers at the King Abdulaziz University in Saudi Arabia have successfully developed and used this methodology to predict the experimental parameters required to generate size-tunable particles with specific characteristics using a microfluidics approach. The initial proof-of-concept study used manufacturer-supplied data (Dolomite Microfluidics) to show that AI could be combined with microfluidics to accelerate research. The aim now is to expand this approach to provide researchers with a tool that enables the acceleration of particle development for a variety of applications, including biomimetic studies, biomedicine and pharmaceuticals.  

Summary

The microfluidics sector is facing an ever-increasing demand to deliver as a platform for polymeric particle manufacturing, and there is a need to have quick and robust high throughput protocols for nanoparticle generation in place. This has led scientists to reach out to computerised tools that can help in bypassing some of the more laborious steps in particle production. Two different platforms –CFD and AI – that can provide incredibly precise simulations and outcomes have now been tested and validated, saving time and reagents, and reducing manual labor. These exciting new technologies can provide scientists with a range of operating protocols that give them a head start in the lab, helping to accelerate microfluidic nanoparticle synthesis, which is a major step forward in the right direction.