Rational payload design to address drug resistance
Background
Antibody-drug conjugates (ADCs) release cytotoxic payloads within target cells through internalization and lysosomal metabolism, enabling selective tumor cell killing.
Role of Efflux Transporters: Cell membrane efflux transporters, such as P-glycoprotein (P-gp) and breast cancer resistance protein (BCRP), reduce intracellular drug concentrations, contributing to drug resistance in oncology therapies.
Limitations in Current Models: Existing P-gp prediction models are general and lack specific data on the Camptothecin (CPT) series. Although BCRP-mediated efflux of CPTs has been reported with systematic SAR analysis, insights into CPT interactions with P-gp remain limited.
Approach
We leveraged Uni-QSAR, a pre-trained model encoded with chemical syntax and structure-related properties.
BCRP-related dataset (~2800 datapoints curated from public sources) was used for training and ~20 camptothecin-specfic dataset was used for fine-tuning.
Biased SAR interpretaion in forms of binding geometry was included during training.
Results
Superior classification performance (>80% accuracy) from the fine-tuned model.
