Skip to main page content
U.S. flag

An official website of the United States government

Dot gov

The .gov means it’s official.
Federal government websites often end in .gov or .mil. Before sharing sensitive information, make sure you’re on a federal government site.

Https

The site is secure.
The https:// ensures that you are connecting to the official website and that any information you provide is encrypted and transmitted securely.

Access keys NCBI Homepage MyNCBI Homepage Main Content Main Navigation
. 2024 Jan-Dec;16(1):2352887.
doi: 10.1080/19420862.2024.2352887. Epub 2024 May 14.

Predicting the clinical subcutaneous absorption rate constant of monoclonal antibodies using only the primary sequence: a machine learning approach

Ronghua Bei  1 , Justin Thomas  1 , Shiven Kapur  1 , Mahlet Woldeyes  1 , Adam Rauk  1 , Jason Robarge  1 , Jiangyan Feng  1 , Kaoutar Abbou Oucherif  1
Affiliations

Predicting the clinical subcutaneous absorption rate constant of monoclonal antibodies using only the primary sequence: a machine learning approach

Ronghua Bei et al. MAbs. 2024 Jan-Dec.

Abstract

Subcutaneous injections are an increasingly prevalent route of administration for delivering biological therapies including monoclonal antibodies (mAbs). Compared with intravenous delivery, subcutaneous injections reduce administration costs, shorten the administration time, and are strongly preferred from a patient experience point of view. An understanding of the absorption process of a mAb from the injection site to the systemic circulation is critical to the process of subcutaneous mAb formulation development. In this study, we built a model to predict the absorption rate constant (ka), which denotes how fast a mAb is absorbed from the site of administration. Once trained, our model (enabled by the XGBoost algorithm in machine learning) can predict the ka of a mAb following a subcutaneous injection using in silico molecular properties alone (generated from the primary sequence). Our model does not need clinically observed plasma concentration-time data; this is a novel capability not previously achieved in predictive pharmacokinetic models. The model also showed improved performance when benchmarked against a recently reported mechanistic model that relied on clinical data to predict subcutaneous absorption of mAbs. We further interpreted the model to understand which molecular properties affect the absorption rate and showed that our findings are consistent with previous studies evaluating subcutaneous absorption through direct experimentation. Taken altogether, this study reports the development, validation, benchmarking, and interpretation of a model that can predict the clinical ka of a mAb using its primary sequence as the only input.

Keywords: Machine learning; Xgboost; monoclonal antibody; rate constant; subcutaneous absorption.

PubMed Disclaimer

Conflict of interest statement

No potential conflict of interest was reported by the author(s).

Figures

Figure 1.
Figure 1.
SC absorption rate constant overview (1a), across IgG subclasses (1b), and across clearance types (1c). For clarity, the two box plots show the 25%, 50%, and 75% quartiles with jittered points.
Figure 2.
Figure 2.
Comparison of the predicted and actual ka values pooled from each cross validation run. The two dotted lines are one standard deviation away from the dashed identity line, and the enclosed area is highlighted in green. Each mAb has 10 runs of corresponding predictions shown as faded, red dots. Their means are shown as blue dots.
Figure 3.
Figure 3.
SHAP plot of the XGBoost model. It is a visual representation of how different molecular property values (also customarily called feature values in a SHAP plot) affect ka prediction. Each row of the plot corresponds to a molecular property and is dotted with 310 points (corresponding to 310 runs obtained from 31 mAbs each run 10 times). The range of values for a given molecular property is color coded from purple (relatively high value for a given molecular property) to yellow (relatively low value for a given molecular property). The position of a colored point relative to the centerline shows the impact of a value. If a point lies to the right of the centerline (and thus with a positive SHAP value), it makes the model predict a higher ka relative to an initial guess (starting at the ka average across the 31 mAbs), and vice versa. For example, if yellow points cluster far away to the right of the center line, the interpretation will be that lower values of a given molecular property is likely related to a higher ka. The magnitude of the SHAP values averaged across all data points for each molecular property is listed by the y-axis. Properties with an averaged SHAP value below or equal to 0.05 (unitless as ka is normalized during data preprocessing) are considered unimportant, which, if converted back, is equivalent to the inability to change more than 0.0002 h−1 from an initial guess. Considering that the ka average across the 31 mAbs is two orders of magnitude larger and the smallest ka is still one order of magnitude larger than 0.0002 h−1, we consider it reasonable to exclude any molecular properties with below or equal to a SHAP value of 0.05 as unimportant.
See this image and copyright information in PMC

Similar articles

See all similar articles

References

    1. Swartz M. The physiology of the lymphatic system. Adv Drug Deliv Rev. 2001;50(1–2):3–11. doi:10.1016/S0169-409X(01)00150-8. - DOI - PubMed
    1. Supersaxo A, Hein WR, Steffen H.. Effect of molecular weight on the lymphatic absorption of water-soluble compounds following subcutaneous administration. Pharm Res. 1990;7(2):167–69. doi:10.1023/A:1015880819328. - DOI - PubMed
    1. Gabrielsson J, Weiner D, Weiner D. Pharmacokinetic & pharmacodynamic data analysis: concepts and applications. In: Gabrielsson J, editors. Expanded.; PK/PD data analysis: concepts and applications; Apotekarsocieteten, Swedish Acedemy of Pharmaceutical Sciences. Stockholm: Swedish Pharmaceutical Press; 2006. p. 43.
    1. Shah DK, Betts AM. Towards a platform PBPK model to characterize the plasma and tissue disposition of monoclonal antibodies in preclinical species and human. J Pharmacokinet Pharmacodyn. 2012;39(1):67–86. doi:10.1007/s10928-011-9232-2. - DOI - PubMed
    1. Cao Y, Jusko WJ. Applications of minimal physiologically-based pharmacokinetic models. J Pharmacokinet Pharmacodyn. 2012;39(6):711–23. doi:10.1007/s10928-012-9280-2. - DOI - PMC - PubMed

Grants and funding

This study was entirely funded by Eli Lilly & Company.

LinkOut - more resources