Transcriptomic and neuroimaging data integration enhances machine learning classification of schizophrenia.

Area of Science:

• Neuroscience
• Genetics
• Machine Learning

Background:

• Schizophrenia is a complex polygenic disorder impacting brain structure and function.
• Integrating macroscale brain features with microscale genetic data offers a comprehensive view of schizophrenia etiology.
• Such integration may yield potential diagnostic markers for schizophrenia.

Purpose of the Study:

• To systematically evaluate the efficacy of fusing multi-scale neuroimaging and transcriptomic data for schizophrenia classification.
• To assess the performance of machine learning models utilizing integrated multi-omics data for schizophrenia diagnosis.

Main Methods:

• Collected brain imaging and blood RNA sequencing data from 43 schizophrenia patients and 60 healthy controls.
• Extracted multi-omics features including macroscale brain morphology, structural/functional connectivity, and gene transcription of schizophrenia risk genes.
• Applied a machine learning integration framework for multi-scale data fusion and patient classification using conventional methods and neural networks.

Main Results:

• Multi-omics data fusion in conventional machine learning models achieved high accuracy (AUC 0.76-0.92), outperforming single-modality models by 8.88-22.64%.
• Multimodal classification models using neural networks showed a 16.57% increase in accuracy (71.43%) compared to single-modal averages.
• Identified key brain regions, including the left posterior cingulate and right frontal pole, crucial for disease classification.

Conclusions:

• Provided empirical evidence for improved schizophrenia classification accuracy through integrated imaging and genetic data.
• Multi-scale data fusion demonstrates significant potential for enhancing diagnostic precision in schizophrenia.
• This approach may facilitate earlier detection and personalized treatment strategies for schizophrenia patients.