Scholarlink

<p>Radiogenomics combines imaging and genomic data to provide insights into disease mechanisms. In this study, we analyzed data from 38,844 UK Biobank participants who had both abdominal MRI scans and genome sequences. Through a genome-wide association study (GWAS), we identified both known and novel single nucleotide polymorphisms (SNPs) associated with liver and pancreas traits. For example, we replicated established associations such as rs1800562 (HFE) with liver iron and rs738409 (PNPLA3) with liver fat. We also discovered potentially novel associations, including rs3734626 near <i>RSPO3</i>, linked to pancreas volume, which to our knowledge has not been previously reported in this context. We then used survival analysis and machine learning to assess how well radiomics, genetic (SNP), and proteomics data could predict liver and pancreas disorders. Key predictors of disease progression included liver iron, pancreas fat content (PDFF), and several SNPs. Survival curves showed clear differences based on these features. Our machine learning models performed well, with the best model (combining all data types) reaching an accuracy of 0.86 —outperforming models using only one data type. These findings highlight the power of integrating imaging, genomic, and proteomic data to improve disease prediction and support personalized healthcare.</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

Multimodal radiogenomic integration enhances prognostic and predictive modeling of liver and pancreas disorders

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年2月3日

查看全文
  • Mohamad Koohi-Moghadam,
  • Varut Vardhanabhuti,
  • Junwen Wang,
  • Albert Chi-Yan Chan,
  • Kyongtae Tyler Bae

摘要

Radiogenomics combines imaging and genomic data to provide insights into disease mechanisms. In this study, we analyzed data from 38,844 UK Biobank participants who had both abdominal MRI scans and genome sequences. Through a genome-wide association study (GWAS), we identified both known and novel single nucleotide polymorphisms (SNPs) associated with liver and pancreas traits. For example, we replicated established associations such as rs1800562 (HFE) with liver iron and rs738409 (PNPLA3) with liver fat. We also discovered potentially novel associations, including rs3734626 near RSPO3, linked to pancreas volume, which to our knowledge has not been previously reported in this context. We then used survival analysis and machine learning to assess how well radiomics, genetic (SNP), and proteomics data could predict liver and pancreas disorders. Key predictors of disease progression included liver iron, pancreas fat content (PDFF), and several SNPs. Survival curves showed clear differences based on these features. Our machine learning models performed well, with the best model (combining all data types) reaching an accuracy of 0.86 —outperforming models using only one data type. These findings highlight the power of integrating imaging, genomic, and proteomic data to improve disease prediction and support personalized healthcare.