Antibody Viscosity Prediction
1 Introduction
High-concentration antibody solutions are essential for the development of subcutaneous injectable formulations, but they often exhibit high viscosity, which poses challenges to antibody drug development, production, and administration. Previous computational models have been limited to training on only a few dozen data points, which is a bottleneck for generalization.
In the DeepViscosity1 work, an ensemble of 102 neural network models was trained using a total of 30 features extracted from the DeepSP model2, including features related to charge distribution, hydrophobicity, spatial aggregation propensity (SAP), and spatial charge map (SCM), to classify low viscosity (≤20cP) and high viscosity (>20cP) mAbs at a concentration of 150mg/mL. Two independent test sets were used to evaluate the generalization capability of DeepViscosity. The model achieved an accuracy of 87.5% and 89.5% on these test sets, respectively, outperforming other prediction methods. This demonstrates that DeepViscosity can facilitate early-stage antibody development by enabling the selection of low-viscosity antibodies, thereby improving developability.
Figure 1. DeepViscosity performance on test sets.
2 Parameters
| Name | Description |
|---|---|
| heavy chain | Heavy chain of the antibody |
| light chain | Light chain of the antibody |
3 Result Description
The result is a .csv table file, with each column defined as follows:
| Name | Description |
|---|---|
| Prob_Mean | Mean of predictions from all models |
| Prob_Std | Standard deviation of predictions from all models |
| DeepViscosity_classes | High viscosity (>20cP) if Prob_Mean ≥ 0.5, otherwise low viscosity (≤20cP) |
4 References
[1] Kalejaye, L. A., Chu, J. M., Wu, I. E., Amofah, B., Lee, A., Hutchinson, M., … Lai, P. K. (2025). Accelerating high-concentration monoclonal antibody development with large-scale viscosity data and ensemble deep learning. mAbs, 17(1). https://doi.org/10.1080/19420862.2025.2483944
[1] Kalejaye, L.; Wu, I.-E.; Terry, T.; Lai, P.-K. DeepSP: Deep Learning-Based Spatial Properties to Predict Monoclonal Antibody Stability. Comput. Struct. Biotechnol. J. 2024, 23, 2220–2229. https://doi.org/10.1016/j.csbj.2024.05.029