Sign in Subscribe
Machine learning

Predicting Subcutaneous Antibody Bioavailability Using Ensemble Protein Language Models

Strategic Insights

— 1 min read

Researchers at Stevens Institute of Technology have made a breakthrough in predicting the subcutaneous bioavailability of monoclonal antibodies, which are crucial in modern therapeutics. According to a study published in Molecular Pharmaceutics, the team developed a novel machine learning framework that leverages protein language models (PLMs) to derive high-dimensional embeddings directly from antibody sequences. This innovative approach has the potential to accelerate the therapeutic development pipeline by enhancing the predictive accuracy and scalability of mAb bioavailability assessment.

Key Takeaways:

  • The researchers introduced a novel machine learning framework that leverages protein language models (PLMs) to derive high-dimensional embeddings directly from antibody sequences.
  • The framework uses three distinct PLMs (antiBERTy, ABlang, and ESM-2) to extract numerical representations that are subsequently refined through feature selection and dimensionality reduction.
  • A systematic evaluation of multiple classifiers using Leave-One-Out cross-validation led to the development of a robust ensemble model based on a tuned support vector machine classifier, which achieved a validation accuracy of 89%.
  • The ensemble approach aggregates predictions across antibodies and outperforms prior computational methods.
  • The researchers deployed the model as a web application, SubQAvail, enabling rapid bioavailability predictions from input antibody sequences.
  • The study demonstrates the potential of integrating PLM-derived features with ensemble learning to enhance the predictive accuracy and scalability of mAb bioavailability assessment.

Statistics:

  • 89% validation accuracy of the tuned support vector machine classifier.
  • 3 distinct PLMs (antiBERTy, ABlang, and ESM-2) used in the framework.
  • 1 robust ensemble model developed based on the tuned support vector machine classifier.
  • 1 web application (SubQAvail) deployed to enable rapid bioavailability predictions.

Sources:

  • Stevens Institute of Technology
  • Molecular Pharmaceutics
  • Amer Chemical Soc (American Chemical Society)
  • NewsRx LLC
  • William Hojegian, Dept. of Chemical Engineering and Materials Science, Stevens Institute of Technology
  • Miles Cabreza, I-En Wu, and Pin-Kuang Lai

Read more