Predicting Drug-Induced Liver Injury Using Machine Learning on a Diverse Set of Predictors
Temidayo Adeluwa1, Brett A McGregor1, Kai Guo1,2
1Department of Biomedical Sciences, University of North Dakota, Grand Forks, ND, United States.
Abstract:
A major challenge in drug development is safety and toxicity concerns due to drug side effects. One such side effect, drug-induced liver injury (DILI), is considered a primary factor in regulatory clearance. The Critical Assessment of Massive Data Analysis (CAMDA) 2020 CMap Drug Safety Challenge goal was to develop prediction models based on gene perturbation of six preselected cell-lines (CMap L1000), extended structural information (MOLD2), toxicity data (TOX21), and FDA reporting of adverse events (FAERS). Four types of DILI classes were targeted, including two clinically relevant scores and two control classifications, designed by the CAMDA organizers. The L1000 gene expression data had variable drug coverage across cell lines with only 247 out of 617 drugs in the study measured in all six cell types. We addressed this coverage issue by using Kru-Bor ranked merging to generate a singular drug expression signature across all six cell lines. These merged signatures were then narrowed down to the top and bottom 100, 250, 500, or 1,000 genes most perturbed by drug treatment. These signatures were subject to feature selection using Fisher's exact test to identify genes predictive of DILI status. Models based solely on expression signatures had varying results for clinical DILI subtypes with an accuracy ranging from 0.49 to 0.67 and Matthews Correlation Coefficient (MCC) values ranging from -0.03 to 0.1. Models built using FAERS, MOLD2, and TOX21 also had similar results in predicting clinical DILI scores with accuracy ranging from 0.56 to 0.67 with MCC scores ranging from 0.12 to 0.36. To incorporate these various data types with expression-based models, we utilized soft, hard, and weighted ensemble voting methods using the top three performing models for each DILI classification. These voting models achieved a balanced accuracy up to 0.54 and 0.60 for the clinically relevant DILI subtypes. Overall, from our experiment, traditional machine learning approaches may not be optimal as a classification method for the current data.
Insights
Developing predictive models for drug-induced liver injury (DILI) faces challenges with gene expression data. Ensemble methods combining multiple data sources showed limited improvement, suggesting traditional machine learning may not be optimal for DILI prediction.
Area of Science:
- Computational biology and cheminformatics
- Toxicology and drug safety assessment
- Biomedical data science
Background:
- Drug-induced liver injury (DILI) is a significant hurdle in drug development, impacting regulatory approval.
- Predicting DILI necessitates integrating diverse data types, including gene expression, structural information, and adverse event reports.
- The Critical Assessment of Massive Data Analysis (CAMDA) 2020 CMap Drug Safety Challenge aimed to advance DILI prediction models.
Purpose of the Study:
- To develop and evaluate machine learning models for predicting DILI using gene perturbation data, structural information, toxicity data, and adverse event reports.
- To address challenges in gene expression data coverage across cell lines by implementing a novel merging strategy.
- To assess the efficacy of traditional machine learning approaches and ensemble methods in DILI classification.
Main Methods:
- Utilized CMap L1000 gene expression data, MOLD2 structural information, TOX21 toxicity data, and FAERS adverse event reports.
- Developed a Kru-Bor ranked merging method to create unified drug expression signatures from six cell lines.
- Applied Fisher's exact test for feature selection and employed ensemble voting (soft, hard, weighted) with top-performing models.
Main Results:
- Models based solely on gene expression signatures showed variable performance (accuracy 0.49-0.67, MCC -0.03-0.1).
- Models using FAERS, MOLD2, and TOX21 data achieved similar predictive results (accuracy 0.56-0.67, MCC 0.12-0.36).
- Ensemble voting models improved balanced accuracy to 0.54 and 0.60 for clinically relevant DILI subtypes, but overall performance remained limited.
Conclusions:
- Traditional machine learning methods may not be optimal for DILI classification with the current dataset.
- Integrating diverse data sources through ensemble methods offers marginal improvements.
- Further research into advanced computational approaches is warranted for accurate DILI prediction.
Related Concept Videos
01:22Effect of Hepatic Disease on Pharmacokinetics: Pathophysiologic Assessment and Liver Function Test
01:26Effect of Hepatic Disease on Pharmacokinetics: Drug Dosing and Hepatic Blood Flow
01:14Effect of Hepatic Disease on Pharmacokinetics: Active Drug, Metabolite and Fraction of Metabolized Drug
01:07Physiological Pharmacokinetic Models: Incorporating Hepatic Transporter-Mediated Clearance
A recent model describes pravastatin's hepatobiliary excretion,...
01:08Effect of Hepatic Disease on Pharmacokinetics: Dose Adjustments Due to Hepatic Impairment
01:16Hepatic Drug Excretion: Influencing Factors