Ensemble learning for higher diagnostic precision in schizophrenia using peripheral blood gene expression profile

Abstract

The need for molecular biomarkers for schizophrenia has been well recognized. Recently, peripheral blood gene expression profiling and machine learning (ML) tools have become popular for biomarker discovery. The stigmatization associated with schizophrenia advocates the need for diagnostic models with higher precision. In this study, we propose a strategy to develop higher-precision ML models using ensemble learning. We performed a meta-analysis using peripheral blood expression microarray data. The ML models: Support Vector Machines (SVM), and Prediction Analysis for Microarrays (PAM) were developed using differentially expressed genes as features. The ensemble of SVM-radial and PAM predicted test samples with a precision of 81.33% (SD: 0.078). The precision of the ensemble model was significantly higher than SVM-radial (63.83%, SD: 0.081) and PAM (66.89%, SD: 0.097). The feature genes identified were enriched for biological processes such as response to stress, response to stimulus, regulation of the immune system, and metabolism of organic nitrogen compounds. The network analysis of feature genes identified PRF1, GZMB, IL2RB, ITGAL, and IL2RG as hub genes. Additionally, the ensemble model developed using microarray data classified the RNA-Sequencing samples with moderately high precision. The pipeline developed in this study allows the prediction of a single microarray and RNA-Sequencing sample. In summary, this study developed robust models for clinical application and suggested ensemble learning for higher diagnostic precision in psychiatric disorders.

A Diagnostic Gene-Expression Signature in Fibroblasts of Amyotrophic Lateral Sclerosis

“…During the last few years, several studies have demonstrated the value and utility of transcriptome profiling of post-mortem tissues in unravelling pathophysiological mechanisms underlying ALS and supporting the existence of a molecular taxonomy for this disease [9,[19][20][21][22][23][24][25][26][27][28][29][30][31][32][33][34]. However, post-mortem analysis of ALS brain samples does not allow for evaluation of alterations occurring during the disease course; thus, it does not represent the optimal resource for biomarker discovery efforts.…”

“…In particular, gene expression studies have been demonstrated to be powerful in providing valuable insights into the molecular basis underlying ALS pathophysiology and identifying molecular signatures or biomarkers able to classify ALS patients into selective clinically relevant subtypes characterized by different biological properties, prognostic biomarkers, and treatment options [14][15][16][17][18][19]. Within this context, the emerging use of machine learning approaches to find genetic biomarkers or construct robust disease classifiers based on patients' gene expression data is revolutionizing clinical decision-making in multiple complex human conditions, including cancer and cardiovascular diseases, and proving to be an exciting tool and promising option for hopefully improving our skills also in neurological conditions [20][21][22][23][24][25][26][27][28].…”