BB-EIT: A Generalized Prediction Model for Protein Adsorption on Polymer Brushes Using Augmented Chemical Embeddings

Area of Science:

• Materials Science
• Biomaterials Engineering
• Computational Chemistry

Background:

• Precise control of protein adsorption on polymer surfaces is crucial for biomaterial design and applications like biosensors and drug delivery.
• Current predictive models struggle with the complexity and diversity of polymer-protein interactions, limiting their generalizability.

Purpose of the Study:

• To develop a generalized computational model for accurately predicting diverse protein adsorption on polymer brushes.
• To overcome limitations in existing models for polymer-protein interaction prediction.

Main Methods:

• Introduced BB-EIT (Biointerface BERT Encoder for Interaction Translation), a novel model based on the ChemBERTa large language model (LLM).
• Utilized SMILES strings for chemical representation and data augmentation.
• Integrated physicochemical and biochemical features (e.g., polymer thickness, protein pI) into an extended model layer.

Main Results:

• BB-EIT demonstrated state-of-the-art performance and strong generalizability in predicting protein adsorption.
• The model accurately predicted adsorption behavior in previously unseen polymer and protein systems.
• Achieved high accuracy in predicting protein adsorption quantities.

Conclusions:

• BB-EIT offers a powerful, generalized approach for predicting protein adsorption on polymer surfaces.
• This work facilitates the data-driven design of advanced biomaterials with tailored interfacial properties.
• Represents a significant advancement in polymer informatics for biomaterial development.