Using multi-trait polygenic scores to predict lithium responsiveness in patients with bipolar disorder.

Area of Science:

• Genetics
• Psychiatry
• Pharmacogenomics

Background:

• Predicting lithium treatment response in bipolar disorder (BD) using polygenic scores (PGSs) has shown limited efficacy.
• Identifying genetic markers for lithium responsiveness is crucial for personalized medicine in BD treatment.

Purpose of the Study:

• To develop and validate a multi-trait polygenic score (mt-PGS) for enhanced prediction of lithium responsiveness in BD patients.
• To integrate genetic information from diverse phenotypes associated with lithium response and BD etiology.

Main Methods:

• Utilized data from the International Consortium on Lithium Genetics (ConLi+Gen) study (N=2,367) including BD patients treated with lithium.
• Calculated PGSs for 59 phenotypes across five clusters (clinical, cardiometabolic, autoimmune/inflammatory, neurocognitive, renal).
• Employed cross-validated machine learning regression to construct mt-PGS models and assessed performance using explained variance (R2) and classification metrics.

Main Results:

• The developed mt-PGS explained 5.07% (continuous) to 9.02% (categorical) of the variability in lithium responsiveness.
• Achieved a classification accuracy (AUC) of 68.13% for predicting favorable lithium response.
• PGSs for clinical lithium exemplar phenotypes showed the strongest association with lithium responsiveness.

Conclusions:

• Integrating PGSs from multiple phenotypes substantially improves prediction of lithium response in BD, offering up to a nine-fold increase in accuracy.
• Future research should focus on larger, diverse cohorts and integrating genetic scores with clinical data for improved clinical implementation.