Browser does not support script.
Professor Mischa Zelzer is a Professor of Drug Delivery in the School of Pharmacy, where he uses artificial intelligence, machine learning and data science to develop predictive models for pharmaceutical formulation.
Professor Mischa Zelzer is a Professor of Drug Delivery in the School of Pharmacy, where he works closely with academic and industry partners to accelerate the development of medicines while reducing costs, material use and environmental impact. In this Take 10, Mischa shares how data-driven approaches are transforming drug formulation, the challenges and opportunities shaping the field, and the experiences that have influenced his research career.
I am working with data to build models that will let us understand better how to adapt or design new materials. I apply this to pharmaceutical research, where the models and breakthroughs we create using artificial intelligence (AI) and machine learning (ML) are used to inform how a drug can be formulated to deliver the right quantity at the right time. As drugs can be expensive and formulation development time-consuming, we also apply our computational approaches to reduce the number of experiments needed to get the right medicine. We do this by creating our own data from materials that we make ourselves such as polymers, gels and lipids and by collaborating with colleagues across academia and industry.
I enjoy working with data because there is always so much to discover. It is thrilling to find new information from a pile of data that was not obvious at first glance and it is even more exciting if that new information can make a difference. Working with a range of colleagues from different disciplines, discussing their challenges, data and insights with them inspires the development of new ideas and approaches.
In the pharmaceutical sector, developing predictive models that explain connections or let us find a new formulation faster is such a powerful aspect that can drive us towards faster and cheaper medicines, medicines that are better adapted to individual patients and of course, new discoveries. It is a privilege to be a part of that community.
With my work, I am making a difference by helping bring machine learning and active learning approaches to the drug formulation space. Drug discovery and clinical trials are important parts of medicines development processes, which have also benefited from AI/ML already. Between these sits the formulation space, which is less evolved in terms of AI/ML applications. Our work fills that gap by bringing in new ideas of how to handle situations where data is disorganised or where little data is available.
Making drug formulation faster and using less material for drug formulation development is a key driver for our work as this will lead to more targeted and efficient patient healthcare, reduce the carbon footprint of drug development and, of course, reduce development cost and time for industry partners.
The majority of my projects involve partners from industry, including small and large pharmaceutical companies. This is important because it fosters a two-way exchange. Colleagues from industry keep our understanding of the problems in formulation science grounded, ensuring we remain conscious of both patient and industrial needs. Our academic exploration helps develop new ideas, find new avenues or demonstrate the feasibility of our breakthroughs to industrial partners, who can then take more informed decisions about adopting them.
While not new, the big questions that remain unsolved in my area are how to deal with inconsistencies in and small sizes of available datasets. While there is a lot of historic data in drug formulation, they are not all reported in the same way. In some areas, more data is available than in others. When generating your own data, not all areas lend themselves to generating large amounts of data, either because available methods are too slow or the amount of material needed is too large and costly. The challenge we face is to develop better methods requiring less material, creating consistent data faster, and developing computational approaches to improve handling of non-ideal datasets. In addition, we require a better understanding of what data we need, as not all data is helpful and getting too much or the wrong data can be just as problematic.
The challenges in the area can be tackled from multiple directions. We need better experimental methods to make and characterise a range of different formulations. We need better computational methods to handle inconsistent and sparse datasets. And we need communities to come together to agree on data standards in the field to help with making our data reusable for different purposes. We also need to bring people together, including academia, industry, regulators and policy makers to develop future proof strategies on how to handle issues like data ownership, governance and compliance.
Finally, we need to better understand and communicate the limitations of predictive approaches to ensure appropriate development and use.
To me, working in research cannot be an individual endeavour. A key challenge was to find the right people that help drive my ideas and vision forward. Working in a team that provides support and opportunities for reflection is critical.
The other important part in academic life is to find time to think. While it may sound strange, it is more difficult to find thinking time as an academic than one might assume. Being conscious about making time to reflect on data, stay updated with developments in the field and develop new ideas is essential to progress a research career, but easily overlooked when planning activities in a diary. Sometimes, I get my thinking time by consciously scheduling it in, other times I get it when pursuing a hobby that allows my mind to wander more freely.
Network early, network a lot, and never stop networking. Progress and impact come from exchanging ideas. Exchange ideas with a variety of people, not just your immediate colleagues. Listen to the challenges in other fields, opportunities often come by moving across traditional boundaries.
It is thrilling to find new information from a pile of data that was not obvious at first glance and even more exciting if that new information can make a difference.