Machine learning prediction of methionine and tryptophan photooxidation susceptibility

Abstract

Photooxidation of methionine (Met) and tryptophan (Trp) residues is common and includes major degradation pathways that often pose a serious threat to the success of therapeutic proteins. Oxidation impacts all steps of protein production, manufacturing, and shelf life. Prediction of oxidation liability as early as possible in development is important because many more candidate drugs are discovered than can be tested experimentally. Undetected oxidation liabilities necessitate expensive and time-consuming remediation strategies in development and may lead to good drugs reaching patients slowly. Conversely, sites mischaracterized as oxidation liabilities could result in overengineering and lead to good drugs never reaching patients. To our knowledge, no predictive model for photooxidation of Met or Trp is currently available. We applied the random forest machine learning algorithm to in-house liquid chromatography-tandem mass spectrometry (LC-MS/MS) datasets (Met, n = 421; Trp, n = 342) of tryptic therapeutic protein peptides to create computational models for Met and Trp photooxidation. We show that our machine learning models predict Met and Trp photooxidation likelihood with 0.926 and 0.860 area under the curve (AUC), respectively, and Met photooxidation rate with a correlation coefficient (Q²) of 0.511 and root-mean-square error (RMSE) of 10.9%. We further identify important physical, chemical, and formulation parameters that influence photooxidation. Improvement of biopharmaceutical liability predictions will result in better, more stable drugs, increasing development throughput, product quality, and likelihood of clinical success.

Keywords: antibody; developability; drug development; mab; machine learning; oxidation; photostability; prediction; stability; therapeutic protein.

Graphical abstract

Figure 1

Stratified cross-validation result comparison (A–C) Cross-validation result comparison between lasso models and random forest models for (A) methionine (Met) photooxidation probability prediction; (B) Met photooxidation rate prediction; and (C) tryptophan (Trp) photooxidation probability prediction. The mean result of 5-fold is plotted and error bars indicate 1 standard deviation. For categorical models, The area under the curve (AUC) is the predictive metric of success, and the coefficient of determination (R²) is used for the regression model.

Figure 2

Regression machine learning model for predicting deamidation rate Predicted Met oxidation abundance (%) was plotted versus the experimental measured oxidation abundance for the independent validation dataset. Individual Mets are plotted as blue circles, and the solid black line indicates where the predicted oxidation level equals the experimental oxidation level. Our regression model predicted the independent set with Q2 of 0.567 and RMSE of 15.5%.

Figure 3

Met categorical, Met regression, and Trp categorical model feature importance Relative importance of each parameter in the categorical model for predicting Met photooxidation probability, Trp photooxidation probability, and Met photooxidation rate was measured for the comparison lasso models and the random forest models. For lasso models, the importance is indicated by the magnitude of each coefficient. For random forest models, the importance was determined by the mean decrease in out-of-bag accuracy when that parameter was excluded from the model.

Figure 4

Met photooxidation pathway Photooxidation of Met occurs, mainly, as a result of interaction with singlet oxygen, yielding the persulfoxide intermediate. At acidic pH, Met persulfoxide can interact with another Met, forming two molecules of Met sulfoxide. At basic pH, formation of Met sulfoxide occurs by reaction with a hydroxide ion, yielding a single Met sulfoxide (+16 Da). Further oxidation of Met sulfoxide results in Met sulfone (+32 Da). Residues are rendered as sticks with Met carbons and sulfurs colored gray and yellow, respectively, and Met photodegradation product carbons and sulfurs colored green and orange, respectively. Oxygen is colored red, and nitrogen is colored blue.

Figure 5

Trp photooxidation pathway Photooxidation of Trp can occur by direct reaction with hydroxyl radicals, yielding hydroxytryptophan (+16 Da), or by reaction with singlet oxygen, yielding n-formylkynurenine (+32 Da) and kynurenine (+4 Da) degradation products. Further oxidation of kynurenine can produce 3-hydroxykynurenine (+20 Da). Residues are rendered as sticks with Trp carbons colored gray and Trp photodegradation product carbons colored green. Oxygen is colored red and nitrogen is colored blue.