Prediction of antibody non-specificity using protein language models and biophysical parameters
Clinical Snapshot
PICO Framework
| P — Population | Human and mouse antibody sequences from public datasets including functional antibodies that bind defined targets |
| I — Intervention | Protein language model (PLM) embeddings and sequence-based biophysical descriptors for predicting antibody non-specificity |
| C — Comparator | Classical machine learning baselines and interpretable biophysical descriptors |
| O — Outcomes | Prediction accuracy of antibody non-specificity, with 10-fold cross-validation accuracy up to 71% for best performing model |
Bottom Line
This computational study demonstrates that protein language models combined with biophysical descriptors can predict antibody non-specificity with moderate accuracy (71%). The research identifies the variable heavy chain domain, particularly H-CDR3, as the primary determinant of non-specific binding. While the prediction accuracy is comparable to classical machine learning approaches, the PLM-based method provides valuable sequence-context representations and complements traditional biophysical analysis. The identification of isoelectric point as a key driver aligns with previous findings. The publicly available models offer a practical tool for early-stage antibody developability assessment, though the moderate accuracy suggests this should supplement rather than replace experimental validation. The approach shows promise for therapeutic antibody and nanobody development, providing a cost-effective screening method that could reduce late-stage development failures due to non-specific binding issues.
Key Findings
P Value: Not reported in abstract
Effect Size: 71% accuracy for best performing VH domain-based ESM-1v LogisticReg model
Primary Outcome: Prediction accuracy of antibody non-specificity using protein language models
Nnt Or Sensitivity: Isoelectric point identified as key biophysical driver of non-specificity
Confidence Interval: Not reported in abstract
Clinical Application
High feasibility with publicly available computational models requiring minimal resources Relevant to Australian biotechnology sector and TGA regulatory submissions for therapeutic antibody development, particularly for early-stage developability assessment Pharmaceutical companies and research institutions developing therapeutic antibodies and nanobodies
Abstract
The development of therapeutic antibodies requires optimizing target binding affinity and pharmacodynamics, while ensuring high developability potential, including minimizing non-specific binding. In this study, we address this problem by predicting antibody non-specificity by two complementary approaches: (1) antibody sequence embeddings by protein language models (PLMs) and (2) a comprehensive set of sequence-based biophysical descriptors. We benchmark the PLM embeddings against interpretable, sequence-derived biophysical descriptors and use fragment-specific models (variable heavy (VH) and variable light (VL) chains, concatenated regions, and individual complementary-determining regions (CDRs) to identify region-level determinants of non-specificity in functional antibodies that bind defined targets. These models were trained on previously published human and mouse antibody data and tested on three public datasets. We show that non-specificity is best predicted from the VH domain and heavy-chain CDRs. These region-level analyses show that heavy-chain features, especially H-CDR3, dominate the predictive signal. The top performing PLM, a VH domain-based Evolutionary Scale Modeling 1 v LogisticReg model, resulted in 10-fold cross-validation accuracy of up to 71%. While predictive accuracy is comparable to classical machine-learning baselines, PLM-based embeddings provide sequence-context representations that yield consistent region-level attribution and provide complementarity to the classical models by enhancing the reliability of predicted non‑specificity scores when used in combination. Our biophysical descriptor-based analysis identified the isoelectric point as a key driver of non-specificity, consistent with previous reports. Our findings highlight the importance of biophysical properties in predicting antibody non-specificity and highlight the potential of PLMs for the development of antibody-based therapeutics. These conclusions are robust to alternative class definitions and are not driven by VH-VL mutational bias. We illustrate the generalizability and practical use of the PLM approach by extending it to therapeutic antibodies and nanobodies, providing a tool for early-stage developability assessment, and we make the models publicly available.
References
- 1.Sakhnini, L. I., Beltrame, L., Fulle, S., Sormanni, P., Henriksen, A., Lorenzen, N., Vendruscolo, M., & Granata, D. (2026). Prediction of antibody non-specificity using protein language models and biophysical parameters. mAbs. https://doi.org/10.1080/19420862.2026.2678000
Related Research
Rheumatic diseases clinics of North America
Machine Learning-Enhanced Autoantibody Discovery and Diagnostics in Systemic Autoimmune Rheumatic Diseases
2 Aug 2026
RMD open
Anakinra for intravenous immunoglobulin-resistant Kawasaki disease: a systematic literature review
31 July 2026
Genome biology
IDEAL-Age: an interpretable deep learning framework for single-cell resolution profiling of immunological aging
27 July 2026
This content is for educational purposes for healthcare professionals only and does not constitute clinical advice. Clinical decisions should be based on individual patient assessment, current guidelines, and appropriate specialist consultation. Editorial Standards · Privacy Policy · Terms of Service