Machine learning driven acceleration of biopharmaceutical formulation development using Excipient Prediction Software (ExPreSo)
1. This paper introduces ExPreSo, a machine learning-driven platform designed to accelerate the selection of excipients for biopharmaceutical formulation development by predicting suitable inactive ingredients based on protein drug properties and product characteristics.
2. ExPreSo uses a dataset of over 350 peptide/protein drug formulations with proven long-term stability, and employs supervised machine learning to predict the presence of excipients, achieving high prediction accuracy for nine prevalent excipients.
3. The model was trained using various features, including protein structural properties and sequence embeddings, and it showed strong predictive power for excipients like acetic acid, histidine, and polysorbate 80, outperforms random predictors with AUCs above 0.7.
4. One of the most impressive aspects of ExPreSo is its speed. The "Fast" version of the model, which uses sequence-based input features, produces predictions in seconds while maintaining comparable predictive performance to slower models that rely on molecular modeling.
5. ExPreSo’s robustness was demonstrated across different formulations and excipient types, with high resilience against biases from platform formulations, proving its utility for excipient selection in both new drug products and reformulations.
6. The system’s ability to predict excipients based on protein sequence data alone, without relying on proprietary formulation details, positions ExPreSo as a versatile and scalable tool for formulation development across the biopharmaceutical industry.
7. The paper emphasizes that ExPreSo’s predictions should serve as a guide for experimental formulation development, helping to reduce the time, costs, and risks associated with excipient screening.
📜Paper:
biorxiv.org/content/10.1101/…
#MachineLearning #Biopharmaceuticals #DrugFormulation #AIinPharma #Excipients #Biotech #ProteinTherapeutics #DeepLearning #PharmaceuticalResearch #FormulationDevelopment