Related Experiment Video
Updated: Oct 27, 2025
08:51
Author Spotlight: Integrated Multi-Omics Analysis for Unveiling Multicellular Immune Signatures in Clinical Heart Attack Cohorts
Published on: September 20, 2024
1.6K
An integrated machine learning framework for a discriminative analysis of schizophrenia using multi-biological data
Peng-Fei Ke1,2,3, Dong-Sheng Xiong1,2,3, Jia-Hui Li1,2,3
1Department of Biomedical Engineering, School of Material Science and Engineering, South China University of Technology, Guangzhou, 510006, Guangdong, China.
Scientific Reports
|July 20, 2021
Summary
Integrating gut microbiota, blood, and electroencephalogram data with machine learning improves schizophrenia diagnosis. Multi-biological data offers superior accuracy for identifying schizophrenia biomarkers.
Area of Science:
- Neuroscience
- Genomics
- Biomarker Discovery
Background:
- Diagnosing schizophrenia effectively remains a significant clinical challenge.
- Objective biomarkers for schizophrenia are lacking, necessitating novel diagnostic approaches.
- Limited research has explored the utility of multi-biological data for schizophrenia diagnosis.
Purpose of the Study:
- To develop and evaluate an integrated machine learning framework for schizophrenia diagnosis using multi-biological data.
- To assess the diagnostic performance of combining gut microbiota, blood, and electroencephalogram (EEG) data.
- To identify key biological features that discriminate schizophrenia patients from healthy controls.
Main Methods:
- A cross-sectional study design was employed.
- Features were extracted from gut microbiota, blood, and EEG data.
- An integrated machine learning framework with classifiers, feature selection, and cross-validation was utilized.
Main Results:
- The support vector machine classifier achieved the highest accuracy (91.7%) and AUC (96.5%) using multi-biological data without feature selection.
- Multi-biological data demonstrated superior discriminative capacity compared to single data types.
- Key discriminative features included specific gut microbiota, blood markers (e.g., superoxide dismutase), and EEG network properties.
Conclusions:
- Integrated multi-biological data analysis shows significant promise for objective schizophrenia diagnosis.
- The developed machine learning framework can aid in understanding schizophrenia pathophysiology.
- This approach may facilitate the development of novel, robust biomarkers for schizophrenia.
