Predicting CYP2D6 phenotype from resting brain perfusion images by gradient boosting.

Area of Science:

• Pharmacogenomics
• Neuroimaging
• Computational Biology

Background:

• Cytochrome P450 2D6 (CYP2D6) is crucial for metabolizing 20% of common drugs, particularly psychotropic medications.
• High CYP2D6 expression in the brain suggests a role in local drug metabolism and potential endogenous functions.
• Genetic variations in CYP2D6 lead to significant individual differences in drug metabolism rates.

Purpose of the Study:

• To investigate the feasibility of predicting CYP2D6 phenotype using functional magnetic resonance imaging (fMRI) resting-state brain perfusion.
• To assess the accuracy of machine learning models in classifying individuals based on their CYP2D6 metabolic status from brain imaging data.

Main Methods:

• Utilized component-wise gradient boosting on fMRI resting-state brain perfusion images from subjects with diverse CYP2D6 genotypes.
• Employed an extended boosting algorithm to enhance model clinical plausibility and interpretability.

Main Results:

• Achieved high sensitivity and specificity (85-87%) for classifying ultrarapid CYP2D6 metabolizers.
• Demonstrated moderate accuracy (71-79%) for classifying poor CYP2D6 metabolizers.
• Identified associations between CYP2D6 genotypic variation and brain regions, notably prefrontal white matter and corpus callosum.

Conclusions:

• Predicting CYP2D6 phenotype from fMRI brain perfusion images is feasible with promising accuracy.
• The developed probabilistic method offers a potential non-invasive alternative to traditional CYP2D6 genotyping.
• Findings support the role of brain structure and function in drug metabolism variability.