Machine Learning for Predicting the Drug-to-Antibody Ratio (DAR) in the Synthesis of Antibody-Drug Conjugates (ADCs)

The pharmaceutical industry faces challenges in developing efficient and cost-effective drug delivery systems. Among various applications, antibody-drug conjugates (ADCs) stand out by combining cytotoxic or bioactive agents with monoclonal antibodies (mAbs) for targeted therapies. However, bioconjugation methods can produce different outcomes, including no bioconjugation, depending on the mAb, the amino acid residues, and the linker-payload (LP) system used. In this work, we developed a machine learning (ML) algorithm capable of predicting bioconjugation outcomes, allowing the design of the best mAb, LP systems, and conditions for the development of efficient ADCs. In particular, we exploited the potential of the XGBoost algorithm in predicting the drug-to-antibody ratio (DAR) in the synthesis of ADCs. Our model demonstrated high predictive accuracy, with R² scores of 0.85 and 0.95 for lysine and cysteine data sets, respectively. The integration of ML algorithms into bioconjugation processes for ADC synthesis offers a promising approach to streamlining ADC development.