Machine learning prediction of antibody aggregation and viscosity for high concentration formulation development of protein therapeutics

Abstract

Machine learning has been recently used to predict therapeutic antibody aggregation rates and viscosity at high concentrations (150 mg/ml). These works focused on commercially available antibodies, which may have been optimized for stability. In this study, we measured accelerated aggregation rates at 45°C and viscosity at 150 mg/ml for 20 preclinical and clinical-stage antibodies. Features obtained from molecular dynamics simulations of the full-length antibody and sequences were used for machine learning model construction. We found a k-nearest neighbors regression model with two features, spatial positive charge map on the CDRH2 and solvent-accessible surface area of hydrophobic residues on the variable fragment, gives the best performance for predicting antibody aggregation rates (r = 0.89). For the viscosity classification model, the model with the highest accuracy is a logistic regression model with two features, spatial negative charge map on the heavy chain variable region and spatial negative charge map on the light chain variable region. The accuracy and the area under precision recall curve of the classification model from validation tests are 0.86 and 0.70, respectively. In addition, we combined data from another 27 commercial mAbs to develop a viscosity predictive model. The best model is a logistic regression model with two features, number of hydrophobic residues on the light chain variable region and net charges on the light chain variable region. The accuracy and the area under precision recall curve of the classification model are 0.85 and 0.6, respectively. The aggregation rates and viscosity models can be used to predict antibody stability to facilitate pharmaceutical development.

DeepSCM: an efficient convolutional neural network surrogate model for the screening of therapeutic antibody viscosity

“…The 61 mAbs viscosity data at 150 mg/mL were from our previous works. 7–9 The first 20 data were preclinical and clinical stage mAbs, and the remaining 41 were commercial mAbs. The solution was at pH 5.5 to pH 6.0 in a 10-20 histidine-HCl buffer.…”

“…The 61 mAbs viscosity data at 150 mg/mL were from our previous works. [7][8][9] The first 20 data were preclinical…”

“…6 Recently, machine learning has been applied to predict high concentration antibody viscosity. [7][8][9] Because of limited experimental data, only conventional machine learning algorithms such as logistic regression, support vector classification, and decision tree classification were applied. The machine learning features relied on domain expertise and published literature.…”

“…The SCM score has also been used as a machine learning feature to predict antibody aggregation. 9,17 The obstacles to applying SCM are the computational cost and difficulties in model construction.…”

“…The calculation of the SCM score requires MD simulations. The SCM model was compared with the machine learning model for the 14 IgG1/14 IgG4P commercial mAbs 8 and 20 preclinical/clinical mAbs 9 mentioned earlier to evaluate the prediction accuracy. The accuracy for the 14 IgG1 and 14 IgG4 commercial mAbs were 0.93 and 0.79, respectively.…”