Scholarlink

<p>Predicting Alzheimer’s Disease (AD) dementia conversion from mild cognitive impairment (MCI) is crucial for therapeutic strategies. Plasma proteomics offers a powerful approach for biomarker identification and predictive modeling. Using SomaScan plasma proteomics in F.ACE cohort, we identified 77 somamers significantly associated with dementia conversion, encompassing immune, inflammatory and neurological processes. Thirteen were replicated in EMIF-AD MBD validation cohorts, including SMOC1. Leveraging machine learning techniques, we developed an optimal model, integrating demographics and 48 proteins, to predict near-future dementia conversion from MCI. It achieved a concordance index of 0.69 and a median time-dependent AUC of 0.75, with comparative performance in A+/T+ subgroups. The 48-protein panel also showed comparative performance with CSF pTau-181. Model can effectively stratify patients by inferred risk. Independent validation evaluated its generalizability, with variant performance decline. These findings demonstrate plasma proteomics’ potential for biomarker discovery and risk prediction, facilitating precision strategies for drug development.</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

Plasma proteomic signatures associate with near-future Alzheimer’s disease dementia conversion in mild cognitive impairment patients

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年1月9日

查看全文
  • Jingyue Xi,
  • Karen Y. He,
  • Liping Hou,
  • Bart Smets,
  • Silke Miller,
  • Ziad S. Saad,
  • Christopher D. Whelan,
  • Ruiz Laza Agustin,
  • Raquel Puerta,
  • Amanda Cano,
  • Cabrera-Socorro Alfredo,
  • Shuwei Li,
  • Yanfei Zhang

摘要

Predicting Alzheimer’s Disease (AD) dementia conversion from mild cognitive impairment (MCI) is crucial for therapeutic strategies. Plasma proteomics offers a powerful approach for biomarker identification and predictive modeling. Using SomaScan plasma proteomics in F.ACE cohort, we identified 77 somamers significantly associated with dementia conversion, encompassing immune, inflammatory and neurological processes. Thirteen were replicated in EMIF-AD MBD validation cohorts, including SMOC1. Leveraging machine learning techniques, we developed an optimal model, integrating demographics and 48 proteins, to predict near-future dementia conversion from MCI. It achieved a concordance index of 0.69 and a median time-dependent AUC of 0.75, with comparative performance in A+/T+ subgroups. The 48-protein panel also showed comparative performance with CSF pTau-181. Model can effectively stratify patients by inferred risk. Independent validation evaluated its generalizability, with variant performance decline. These findings demonstrate plasma proteomics’ potential for biomarker discovery and risk prediction, facilitating precision strategies for drug development.