Scholarlink

Background <p>Schizophrenia (SCZ) is a highly heritability psychological disorder, however the exact etiology remains unclear, and lack of the reliable and effective biomarkers for diagnosis and treatments in the management of SCZ, thus exploring the novel biomarkers in SCZ may enhance the efficacy of its predictive, preventive, and personalized medicine (PPPM/3PM) approach.</p> Methods <p>Based on differentially expressed genes (DEGs) and weighted gene co-expression network (WGCNA) analyses from five brain datasets, we screened SCZ-key genes and then developed a novel machine learning (ML) framework that incorporated 12 MLs and their 84 combinations to construct a consensus diagnostic signature. Meanwhile, we constructed the nomogram with aforementioned signatures to provide a quantitative clinical practice tool for predicting SCZ. Subsequently, we performed the consensus clustering and nonnegative matrix factorization (NMF) algorithms for clustering analysis in SCZ patients. On this basis, the regulation factors of diagnostic signature, enrichment patterns and immune infiltration analysis in SCZ, and protein level among SCZ subtypes were evaluated.</p> Results <p>We identified 53 SCZ-key genes by intersecting DEGs and module genes of WGCNA, then developed a consensus diagnostic signature using a 84-combination ML framework, and established a nomogram diagnosis model with aforementioned signature for clinical practice, demonstrating promising discriminative performance and potential clinical utility benefits in predicting SCZ. Moreover, consensus clustering analysis could divide SCZ patients into two distinct clusters, and two subgroups were distinguished using NMF algorithm with DEGs of two clusters. Furthermore, we observed distinct biological functions, immune cells and protein functions between subtypes. Finally, hub genes of subgroups, which were closely associated with SCZ.</p> Conclusion <p>Our study constructed a novel diagnostic signature and a nomogram, which all achieved higher accuracy and maybe as the potential diagnostic tools for SCZ. Meanwhile, SCZ subtypes showed distinct inflammation, immune and metabolic patterns, incorporating the subtypes into the 3PM framework will provide a unique opportunity for clinical intelligence and new management approaches.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Machine learning-based integration identifies a 10-gene predictive signature and its classification patterns in schizophrenia

  • 期刊论文

  • 发表日期：

    2026年5月14日

查看全文
  • Yan Li,
  • Qing Sun,
  • Ye Shen,
  • Xinwei Li,
  • Haoyu Li,
  • Jing Ni,
  • Jing Wang,
  • Siyu Sun,
  • Yan Wang,
  • Zhijun Li

摘要

Background

Schizophrenia (SCZ) is a highly heritability psychological disorder, however the exact etiology remains unclear, and lack of the reliable and effective biomarkers for diagnosis and treatments in the management of SCZ, thus exploring the novel biomarkers in SCZ may enhance the efficacy of its predictive, preventive, and personalized medicine (PPPM/3PM) approach.

Methods

Based on differentially expressed genes (DEGs) and weighted gene co-expression network (WGCNA) analyses from five brain datasets, we screened SCZ-key genes and then developed a novel machine learning (ML) framework that incorporated 12 MLs and their 84 combinations to construct a consensus diagnostic signature. Meanwhile, we constructed the nomogram with aforementioned signatures to provide a quantitative clinical practice tool for predicting SCZ. Subsequently, we performed the consensus clustering and nonnegative matrix factorization (NMF) algorithms for clustering analysis in SCZ patients. On this basis, the regulation factors of diagnostic signature, enrichment patterns and immune infiltration analysis in SCZ, and protein level among SCZ subtypes were evaluated.

Results

We identified 53 SCZ-key genes by intersecting DEGs and module genes of WGCNA, then developed a consensus diagnostic signature using a 84-combination ML framework, and established a nomogram diagnosis model with aforementioned signature for clinical practice, demonstrating promising discriminative performance and potential clinical utility benefits in predicting SCZ. Moreover, consensus clustering analysis could divide SCZ patients into two distinct clusters, and two subgroups were distinguished using NMF algorithm with DEGs of two clusters. Furthermore, we observed distinct biological functions, immune cells and protein functions between subtypes. Finally, hub genes of subgroups, which were closely associated with SCZ.

Conclusion

Our study constructed a novel diagnostic signature and a nomogram, which all achieved higher accuracy and maybe as the potential diagnostic tools for SCZ. Meanwhile, SCZ subtypes showed distinct inflammation, immune and metabolic patterns, incorporating the subtypes into the 3PM framework will provide a unique opportunity for clinical intelligence and new management approaches.