Developing a Genetic Biomarker-based Diagnostic Model for Major Depressive Disorder using Random Forests and...

Area of Science:

• Genomics
• Computational Biology
• Psychiatry

Background:

• Current major depressive disorder (MDD) diagnosis relies on subjective assessments.
• There is a need for objective and accurate diagnostic methods for MDD.

Purpose of the Study:

• To develop a novel diagnostic model for predicting MDD using gene expression data.
• To identify reliable gene biomarkers for MDD diagnosis.

Main Methods:

• Utilized human brain and blood gene expression datasets (GSE102556, GSE98793, GSE76826) from the Gene Expression Omnibus (GEO) database.
• Employed a random forest (RF) plus artificial neural network (ANN) algorithm to analyze differentially expressed genes (DEGs).
• Selected and validated 28 gene biomarkers for MDD prediction.

Main Results:

• Identified 100 shared DEGs between brain and blood datasets.
• Developed an RF model that selected 28 predictive gene biomarkers.
• An ANN model achieved high accuracy in predicting MDD, with AUCs of 0.903 and 0.917 for independent datasets.

Conclusions:

• This study presents the first classifier using DEG biomarkers as an endophenotype for MDD diagnosis.
• The findings offer a new avenue for MDD diagnosis, treatment, outcome prediction, prognosis, and recurrence management.