EpiSmokEr

EpiSmokEr predicts smoking status from whole-blood DNA methylation profiles using machine learning to distinguish current and never smokers.

Key Features:

• Machine learning-based classifier: Uses machine learning algorithms to train a classifier that predicts smoking status from DNA methylation data.
• Whole-blood DNA methylation input: Operates on whole-blood DNA methylation profiles as the primary input data type.
• Cross-dataset robustness: Validated across three independent whole-blood datasets and does not require dataset-specific thresholding.
• Phenotypic misclassification evaluation: Includes evaluation of biologically meaningful misclassifications to assess phenotypic influences on methylation patterns.

Scientific Applications:

• Epidemiological studies: Derives methylation-based smoking status to assess smoking prevalence and exposure within populations.
• Cancer research: Identifies smoking-associated DNA methylation changes relevant to studies of smoking-related cancers.
• Personalized medicine: Provides methylation-derived smoking status from biological samples to inform individual risk assessment and stratification.

Methodology:

Analyzes whole-blood DNA methylation data using machine learning algorithms to train and validate a smoking-status classifier across three independent whole-blood datasets and evaluates biologically meaningful misclassifications.