In Silico Formulation Development: What CMC Teams Can Decide Before The Lab

Formulation risks often become visible only during tech transfer, when capital has already been committed, and changes can mean costly delays. Sequence-based prediction offers CMC teams an earlier way to narrow key decisions before lab work begins. By combining machine learning with physics-based simulation, the approach can support candidate selection and identify practical starting ranges for buffer, pH, and excipients. Examples show how three antibody candidates led to three distinct formulation recommendations, while a non-antibody protein prediction was compared with results from an 11-day stability study. The analysis also examines accuracy, false positives, and current coverage limitations beyond antibodies. Rather than replacing experimental work, these predictions can help teams focus testing where it matters most.

Explore the examples to see where computational recommendations aligned with laboratory data.