Integration of machine learning and pharmacogenomic biomarkers for predicting response to antidepressant treatment: can computational intelligence be used to augment clinical assessments?

[...]for many patients with MDD, several therapeutic trials, each lasting several weeks, are needed before a good outcome of treatment with antidepressants is observed (7), thus prolonging disability and suffering in afflicted patients. To be sure, pharmacogenomic research in depression has led to an improved understanding of the neurobiological mechanisms of response to antidepressant treatment and to the discovery of potentially important effect modifiers of therapeutic response and sensitivity to adverse effects (16). [...]a recent meta-analysis of five randomized trials (1737 depressed subjects) showed that individuals who received pharmacogenetics-guided decision support had a significantly higher likelihood of achieving depressive symptom remission than those who received treatment as usual (17). [...]the prognostic outputs and the clinical recommendations that stem from artificial intelligence-based tools may be subject to inadvertent bias with limited or no interpretability of results (31). [...]the thoughtful combination of pharmacogenomics and machine learning holds considerable promise for the goals of achieving the prediction of therapeutic outcomes of antidepressant treatment in patients with MDD with sufficient accuracy for use in the clinic, and for subtyping depressed patients into groups that may be especially likely (or not) to respond to specific treatments.