Practical Conformal Prediction Methods for QSAR

• Activity Number: 166 - Non-Clinical Statistics, Personalized Medicine, and Other Topics
• Type: Contributed
• Date/Time: Tuesday, August 10, 2021 : 10:00 AM to 11:50 AM
• Sponsor: Biopharmaceutical Section
• Author(s): Yuting Xu* and Andy Liaw and Vladimir Svetnik
• Companies: Merck & Co., Inc. and Merck & Co., Inc. and Merck & Co.
• Keywords: Machine Learning; Conformal Prediction; QSAR; Prediction Interval

Abstract

Quantitative structure?activity relationship (QSAR) regression model is a commonly used technique for predicting biological activities of compounds using their molecular descriptors. Predictions from QSAR models help to prioritize compounds for further experimental testing. In addition to accurate predictions, it is highly desirable to obtain some estimate of the prediction uncertainty, such as prediction intervals with a pre-specified nominal coverage probability. The challenge is that most machine learning (ML) algorithms require companion algorithms for estimating uncertainty of their prediction. Development of these algorithms is an active research area in statistical and ML communities but the implementation for QSAR modeling remains limited. Conformal prediction (CP) is a model-agnostic framework that addresses this challenge by producing valid prediction intervals under weak assumptions on data distribution. We proposed computationally efficient CP algorithms tailored to the most advanced ML models, Deep Neural Nets and Gradient boosting. The validity and efficiency of proposed conformal predictors are demonstrated on a diverse collection of QSAR datasets.