A hybrid model for predicting response to risperidone after first-episode psychosis

Area of Science:

• Psychiatry and Pharmacology
• Genetics and Personalized Medicine
• Computational Biology

Background:

• Antipsychotic drug response in first-episode psychosis (FEP) exhibits significant inter-patient variability.
• This heterogeneity may stem from clinical and genetic factors.
• Personalized treatment approaches are needed to optimize outcomes.

Purpose of the Study:

• To evaluate the predictive performance of clinical, genetic, and hybrid models for risperidone response in FEP patients.
• To identify key clinical and genetic predictors of treatment response.

Main Methods:

• 141 antipsychotic-naive FEP patients were assessed before and after 10 weeks of risperidone treatment.
• Response was defined as >=50% reduction in Positive and Negative Syndrome Scale (PANSS) scores.
• Machine learning algorithms (SVM, kNN, RF) were used to build and test clinical, genetic, and hybrid models.

Main Results:

• Clinical models achieved a balanced accuracy of 63.3% (SVM).
• Genetic models showed a balanced accuracy of 58.5% (kNN).
• The best performing hybrid model (RF) integrated clinical data (duration of untreated psychosis, CGI-S, age, cannabis use) and 406 single-nucleotide variants (SNVs), achieving 72.9% balanced accuracy.

Conclusions:

• A hybrid model incorporating both clinical and genetic predictors offers superior prediction of antipsychotic treatment response in FEP.
• This approach holds promise for personalized pharmacotherapy in psychosis.