A Descriptor Set for Quantitative Structure-property Relationship Prediction in Biologics

Area of Science:

• Biochemistry
• Computational Biology
• Drug Development

Background:

• Biologics, particularly monoclonal antibodies, are increasingly prevalent.
• Drug 'developability' (solubility, low aggregation, stability, etc.) is critical for clinical success.
• Late-stage discovery of developability issues poses significant challenges.

Purpose of the Study:

• To develop an in silico method for early detection of protein developability liabilities.
• To create a computational framework for predicting aggregation and solubility issues.
• To aid in improving protein properties during biologics drug development.

Main Methods:

• Developed a quantitative structure-property relationship (QSPR) computational framework.
• Utilized a diverse set of protein-specific descriptors (sequence, structure, surface patches).
• Assessed descriptor performance through four case studies and novel experimental data.

Main Results:

• The computational framework effectively predicts protein developability liabilities.
• Descriptors demonstrated suitability for identifying and addressing poor solubility and aggregation.
• Validated on public datasets and novel experimental data, including Adnectins.

Conclusions:

• The developed in silico framework aids in early identification of developability issues.
• This approach can significantly improve the developability of biologics candidates.
• Facilitates the advancement of biologics with enhanced physico-chemical properties.