Machine Learning approaches for biotherapeutics
Due to their specificity and targeted effects (minimizing unwanted side effects) biologics represent an important class of drugs. Biologics are very large complex molecules manufactured in living system such as microorganisms, animal cells or plant cells. The global biologics market was worth $221 billion in 2017 and is essentially segmented into monoclonal antibodies, therapeutic proteins and vaccines. The product development and process development phase of such molecules are complex, long and challenging. During the product development phase, several candidates are screened and biophysical assays are performed to characterize biological activity and stability. Once a drug is approved, the task is to produce mass quantities at consistent quality to meet the increasing market demand. Thus, determining optimal production condition and controlling minutely the environment of the cells are key aspects of process development. If done only on wet-lab basis, all of these steps are quite time and resource consuming. Computational tools have proven to accelerate the development of biologics are several fronts saving lots of resources.
In our group, the projects focus on using machine learning tools to plan experiments, determine optimal operating conditions and controlling the production of monoclonal antibodies and extracellular vesicles. Additionally, there is also focus on using artificial intelligence in designing the formulation for the biologics to enhance their stability.
Contacts
Key publications