Logo
ScienceStack
Table of Contents
Fetching ...
Sign In

Auto-ADMET: An Effective and Interpretable AutoML Method for Chemical ADMET Property Prediction

Alex G. C. de Sá, David B. Ascher

TL;DR

ADMET prediction in drug discovery faces data drift and a need for personalised, generalisable pipelines. Auto-ADMET introduces an interpretable AutoML framework that combines grammar-based genetic programming with a Bayesian Network Classifier to automatically construct and justify ADMET pipelines on input chemical datasets. Across twelve benchmark datasets, Auto-ADMET delivers strong predictive performance (average MCC around 0.618) on eight tasks relative to pkCSM, XGBoost, and a standard GGP baseline, while also enabling interpretability through BNC-guided guidance. This work demonstrates that coupling grammar-based pipeline search with causal interpretation can yield both effective cheminformatics AutoML solutions and actionable insights for ADMET modelling.

Abstract

Machine learning (ML) has been playing important roles in drug discovery in the past years by providing (pre-)screening tools for prioritising chemical compounds to pass through wet lab experiments. One of the main ML tasks in drug discovery is to build quantitative structure-activity relationship (QSAR) models, associating the molecular structure of chemical compounds with an activity or property. These properties -- including absorption, distribution, metabolism, excretion and toxicity (ADMET) -- are essential to model compound behaviour, activity and interactions in the organism. Although several methods exist, the majority of them do not provide an appropriate model's personalisation, yielding to bias and lack of generalisation to new data since the chemical space usually shifts from application to application. This fact leads to low predictive performance when completely new data is being tested by the model. The area of Automated Machine Learning (AutoML) emerged aiming to solve this issue, outputting tailored ML algorithms to the data at hand. Although an important task, AutoML has not been practically used to assist cheminformatics and computational chemistry researchers often, with just a few works related to the field. To address these challenges, this work introduces Auto-ADMET, an interpretable evolutionary-based AutoML method for chemical ADMET property prediction. Auto-ADMET employs a Grammar-based Genetic Programming (GGP) method with a Bayesian Network Model to achieve comparable or better predictive performance against three alternative methods -- standard GGP method, pkCSM and XGBOOST model -- on 12 benchmark chemical ADMET property prediction datasets. The use of a Bayesian Network model on Auto-ADMET's evolutionary process assisted in both shaping the search procedure and interpreting the causes of its AutoML performance.

Auto-ADMET: An Effective and Interpretable AutoML Method for Chemical ADMET Property Prediction

TL;DR

ADMET prediction in drug discovery faces data drift and a need for personalised, generalisable pipelines. Auto-ADMET introduces an interpretable AutoML framework that combines grammar-based genetic programming with a Bayesian Network Classifier to automatically construct and justify ADMET pipelines on input chemical datasets. Across twelve benchmark datasets, Auto-ADMET delivers strong predictive performance (average MCC around 0.618) on eight tasks relative to pkCSM, XGBoost, and a standard GGP baseline, while also enabling interpretability through BNC-guided guidance. This work demonstrates that coupling grammar-based pipeline search with causal interpretation can yield both effective cheminformatics AutoML solutions and actionable insights for ADMET modelling.

Abstract

Machine learning (ML) has been playing important roles in drug discovery in the past years by providing (pre-)screening tools for prioritising chemical compounds to pass through wet lab experiments. One of the main ML tasks in drug discovery is to build quantitative structure-activity relationship (QSAR) models, associating the molecular structure of chemical compounds with an activity or property. These properties -- including absorption, distribution, metabolism, excretion and toxicity (ADMET) -- are essential to model compound behaviour, activity and interactions in the organism. Although several methods exist, the majority of them do not provide an appropriate model's personalisation, yielding to bias and lack of generalisation to new data since the chemical space usually shifts from application to application. This fact leads to low predictive performance when completely new data is being tested by the model. The area of Automated Machine Learning (AutoML) emerged aiming to solve this issue, outputting tailored ML algorithms to the data at hand. Although an important task, AutoML has not been practically used to assist cheminformatics and computational chemistry researchers often, with just a few works related to the field. To address these challenges, this work introduces Auto-ADMET, an interpretable evolutionary-based AutoML method for chemical ADMET property prediction. Auto-ADMET employs a Grammar-based Genetic Programming (GGP) method with a Bayesian Network Model to achieve comparable or better predictive performance against three alternative methods -- standard GGP method, pkCSM and XGBOOST model -- on 12 benchmark chemical ADMET property prediction datasets. The use of a Bayesian Network model on Auto-ADMET's evolutionary process assisted in both shaping the search procedure and interpreting the causes of its AutoML performance.

Paper Structure

This paper contains 16 sections, 1 equation, 4 figures, 3 tables.

Table of Contents

  1. Introduction
  2. Related Work
  3. Evolutionary Algorithms for Computation Chemistry
  4. AutoML for Cheminformatics
  5. Auto-ADMET: AutoML for ADMET Property Prediction
  6. Search Space
  7. Search Method
  8. Fitness Function
  9. Experiments
  10. Chemical ADMET Datasets
  11. Parameter Configuration
  12. Benchmarking
  13. Results
  14. Benchmarking against Alternative Methods
  15. Auto-ADMET's interpretation through Bayesian Network Analysis
  16. ...and 1 more sections

Figures (4)

  • Figure 1: Auto-ADMET's workflow to create personalised machine learning (ML) pipelines targeting ADMET chemical compound property prediction.
  • Figure 2: The excerpt of the proposed AutoML grammar.
  • Figure 3: The grammar-based genetic programming (GGP) method with a Bayesian Network Classifier to search for ML pipelines in the context of ADMET property prediction. Figure adapted from deSa2017deSa2024.
  • Figure 4: The built Bayesian Network Classifiers across Auto-ADMET's generations.