Application of machine learning in diagnostic value of mRNAs for bipolar disorder.

PURPOSE: Bipolar disorder (BD) is a type of severe mental illness with symptoms of mania or depression, it is necessary to find out effective diagnostic biomarkers for BD due to diagnosing BD is based on clinical interviews without objective indicators.

MATERIALS AND

METHODS

The mRNA expression levels of genes included PIK3R1, FYN, TP53, PRKCZ, PRKCB, and YWHAB in the peripheral blood of 43 patients with bipolar disorder and 47 healthy controls were detected. Machine learning methods included Artificial Neural Networks, Extreme Gradient Boosting, Random Forest, and Support Vector Machine were adopted to fit different gene combinations to evaluate diagnostic value for bipolar disorder.

RESULTS

The combination 'PIK3R1 + FYN' in the SVM model showed the best diagnostic value, with AUC, sensitivity, and specificity values of 0.951, 0.928, and 0.937, respectively.

CONCLUSIONS

The diagnostic efficiency for bipolar disorder was significantly improved by fitting PIK3R1 and FYN through the Support Vector Machine model.