The Power of Computational Predictors: Antibody Developability

Introduction: Why Computational Developability Matters

Before the road towards clinical trials, the developability properties of therapeutic antibodies need to be evaluated. Antibody molecules are assessed for developability characteristics including aggregation, immunogenicity, pharmacokinetic clearance, viscosity, and half-life span.¹ This is important to ensure a stable and efficacious drug before clinical trials.

These properties can be understood through in silico tools, such as homology modelling, docking or interface prediction. During the Lead Identification and Optimization phases these tools can help to researchers to generate three-dimensional models of the antibodies and predict or identify the key residues involved in antigen binding. Machine learning tools can also predict these properties by assessing hydrophobicity, electrostatic charge, and topology pattern interaction of the amino acid sequence.^2,3

In Silico Approaches to Antibody Design

During lead identification and optimization, researchers can employ tools like homology modeling, docking, and interface prediction to generate 3D antibody structures. These models aid in locating residues critical for antigen binding. Additionally, machine learning algorithms evaluate hydrophobicity, electrostatic charge, and topology interactions from amino acid sequences.

Such in silico tools reduce experimental burden and improve screening efficiency during preclinical studies by processing large volumes of developability data.

Key Computational Tools for Antibody Developability

Spatial-Charge-Map (SCM)

Spatial-Charge-Map (SCM) is a tool developed by Agrawal et al. (2016) that can screen monoclonal antibodies for their viscosities. SCM works by quantifying the electronegative potential patch on the Fv domain of antibodies. High viscosities, commonly found with highly concentrated antibody solutions, poses several problems to develop, manufacture, and administer the potential drug. Since the mAb sequence is a component for viscosity, this SCM tool is extremely helpful for the selection of low-viscosity antibodies.⁴

Therapeutic Antibody Profiler (TAP)

The therapeutic antibody profiler (TAP) is an application which compares your antibody variable domain sequence against five factors: total complementarity-determining region (CDR) length, surface hydrophobicity, asymmetry in net heavy- and light-chain surface charges, and positive charge and negative charge in the CDRs. Raybould et al. (2019) published this tool to reveal antibodies that have rare characteristics in clinical-stage mAb therapeutics.^5,6

AlphaFold for Structural Prediction

AlphaFold, developed by DeepMind, is an AI system that predicts the 3D structure of proteins based on their amino acid sequences. It has been used to model antibody-antigen complexes, providing rapid, high-quality structural models that can inform experimental design. AlphaFold enables researchers to bypass challenges associated with traditional methods like X-ray crystallography and cryo-electron microscopy, which are often slow and expensive.

In the context of therapeutic development, AI-driven approaches offer several benefits:

• Structural Optimization: AI tools predict key regions like the complementarity-determining regions (CDRs), allowing researchers to optimize these regions for better specificity and binding affinity.

• Antigen-Antibody Docking: Predicting the interaction between an antibody and its target antigen is crucial for designing antibodies that bind with high precision. AlphaFold has improved the accuracy of antigen-antibody docking models.

• Reducing Experimental Workload: By providing reliable structural predictions, AlphaFold minimize the need for multiple rounds of experimental validation, saving both time and resources in antibody discovery and development.

• Structural Antibody Database: Insights from AlphaFold predictions contribute to building a growing structural antibody database, enhancing collective knowledge and design efficiency.

Machine Learning Tools for Predictive Profiling

Random Forest for Hydrophobicity Prediction

Jain et al. (2017) used a random forest machine learning approach to estimate the hydrophobic chromatography (HIC) retention time directly from an antibody sequence. Quantifying hydrophobicity of a monoclonal antibody is important to assess downstream risks. The researchers successfully developed predictive models to predict the surface exposure of amino-acid side-chains in the antibody’s variable region.⁷

Predicting Developability and Binding Affinity

Machine learning in biotech also aids in predicting developability, encompassing manufacturability, stability, and solubility. ML models help flag sequence variants with poor developability early in the design process, preventing costly downstream failures. Additionally, ML algorithms are used to predict binding affinity based on antibody sequence data, helping refine antibody candidates before conducting labor-intensive binding assays.

Virtual Antibody Libraries

ML can be used to generate virtual antibody libraries, simulating diverse human antibody populations. This enables researchers to select promising candidates in silico before moving to experimental validation, streamlining the discovery process.

Feedback Loops Between Experimental and In Silico Data

Leading-edge platforms now integrate closed-loop systems, using experimental outcomes to continuously refine in silico tools. As antibodies undergo lab testing, the resulting data feeds back into AI models, retraining them to enhance predictive accuracy. This data feedback loop improves screening outcomes not only for human antibodies, but also for complex targets such as conformational or glycosylated antigens. Such integration strengthens the reliability of computational tools and accelerates therapeutic candidate selection.

Supporting Discovery with AI-Based Solutions

Beyond standalone tools, comprehensive platforms such as SAbPred support antibody structure prediction and developability profiling. These resources integrate computational tools with AI-powered screening to guide antibody engineers toward higher success rates.

At Biointron, we are dedicated to accelerating your antibody discovery, optimization, and production needs. Our team of experts can provide customized solutions that meet your specific research needs. Contact us to learn more about our services and how we can help accelerate your research and drug development projects.

References:

1. Raybould, M.I.J., Deane, C.M. (2022). The Therapeutic Antibody Profiler for Computational Developability Assessment. In: Houen, G. (eds) Therapeutic Antibodies. Methods in Molecular Biology, vol 2313. Humana, New York, NY. https://link.springer.com/protocol/10.1007/978-1-0716-1450-1_5

2. Kim, J., McFee, M., Fang, Q., Abdin, O., & Kim, P. M. (2023). Computational and artificial intelligence-based methods for antibody development. Trends in Pharmacological Sciences, 44(3), 175–189. https://www.cell.com/trends/pharmacological-sciences/fulltext/S0165-6147(22)00279-6

3. Mieczkowski, C., Zhang, X., Lee, D., Nguyen, K., Lv, W., Wang, Y., Zhang, Y., Way, J., & Gries, M. (2023). Blueprint for antibody biologics developability. MAbs, 15(1). https://www.tandfonline.com/doi/full/10.1080/19420862.2023.2185924

4. Agrawal, N. J., Helk, B., Kumar, S., Mody, N., Sathish, H. A., Samra, H. S., Buck, P. M., Li, L., & Trout, B. L. (2016). Computational tool for the early screening of monoclonal antibodies for their viscosities. MAbs, 8(1), 43-48. https://www.tandfonline.com/doi/full/10.1080/19420862.2015.1099773

5. J. Raybould, M. I., Marks, C., Krawczyk, K., Taddese, B., Nowak, J., Lewis, A. P., Bujotzek, A., Shi, J., & Deane, C. M. (2019). Five computational developability guidelines for therapeutic antibody profiling. Proceedings of the National Academy of Sciences of the United States of America, 116(10), 4025-4030. https://www.pnas.org/doi/full/10.1073/pnas.1810576116

6. James Dunbar and others, SAbPred: a structure-based antibody prediction server, Nucleic Acids Research, Volume 44, Issue W1, 8 July 2016, Pages W474–W478, https://academic.oup.com/nar/article/44/W1/W474/2499346

7. Jain, T., Boland, T., Lilov, A., Burnina, I., Brown, M., Xu, Y., & Vásquez, M. (2017). Prediction of delayed retention of antibodies in hydrophobic interaction chromatography from sequence using machine learning. Bioinformatics, 33(23), 3758-3766. https://academic.oup.com/bioinformatics/article/33/23/3758/4083264