内容来源：https://www.sciencenews.org/article/experimental-blood-test-dementia-types

内容总结：

新研究：血检可“一针见血”识别四种痴呆症，准确率超92%

一项实验性的血液检测技术有望帮助医生更精准地诊断痴呆症——不仅能区分阿尔茨海默病、帕金森病等四种主要神经退行性疾病，还能以92.3%的准确率判断患者是否同时患有其中两种或以上疾病。相关研究成果5月发表在《阿尔茨海默病与痴呆症》期刊上。

据统计，美国现有超过600万痴呆症患者，该病已成为全球第七大死因。研究负责人、华盛顿大学人类基因组学家卡洛斯·克鲁查加指出：“这些疾病比我们最初想象的更复杂，重叠现象也比预想中更多。要真正理解其生物学机制，必须将它们放在一起研究。”

不同种类的痴呆症需要不同的护理方案，即使症状相似。明确疾病组合有助于为患者制定更精准的个体化治疗。去年，美国食品药品监督管理局（FDA）已批准首个阿尔茨海默病血液检测，但市面上多数检测仅能锁定单一疾病。而这款名为GPND-AI的新型测试通过分析血液中15种蛋白质水平，可同时识别阿尔茨海默病、帕金森病、额颞叶痴呆和路易体痴呆四种疾病。例如，它不仅能告知患者同时患有阿尔茨海默病和帕金森病，还能具体指出：检出的异常蛋白中，75%指向阿尔茨海默病，20%指向帕金森病。

研究团队基于华盛顿大学两家专科诊所的3000余名患者血液样本和病历，借助人工智能算法从123种候选蛋白中筛选出15种关键标志物，其中包括现有阿尔茨海默病血检的核心指标p-tau217。随后，团队在美国亚利桑那州班纳太阳健康研究所的另一组患者中进行验证，将血检结果与患者去世后捐献的脑组织病理分析对比，证实了模型的可靠性。

波士顿塔夫茨医学中心神经心理学家达维德·卡彭评价，该检测“从简单的‘是/否’判断转向反映临床中常见的生物学复杂性”，是领域内“重要的概念性转变”。但他也强调，仍需更大规模、更多样化的研究来进一步验证。

“五六年前，没人相信血液检测能识别阿尔茨海默病或神经退行性疾病，”克鲁查加感叹道，“但技术进步的速度正在加快。”目前，团队正与制药企业洽谈，计划开展临床试验并启动FDA审批流程。卡彭特别提醒，这类血检无法完全替代临床评估：“血检能提示潜在病理，但不能解释患者的日常功能状况、症状起因，以及睡眠、情绪、药物、血管疾病等因素对认知能力的影响。”

中文翻译：

一项针对痴呆症的血液检测，或许能告诉你是否患有不止一种类型
这种实验性检测有助于医生发现重叠的脑部疾病

当痴呆症患者的大脑出现问题时，往往不止一种病因。但要将多种脑部疾病区分开来颇为困难，尤其是在早期阶段，甚至难以判断是否同时存在多种疾病。一项新型实验性血液检测可能会改变这一现状。

该检测通过测量血液中15种蛋白质的水平，来诊断四种主要的神经退行性疾病——阿尔茨海默病、帕金森病、额颞叶痴呆和路易体痴呆。研究人员在5月刊的《阿尔茨海默病与痴呆》杂志中报告称，它能以92.3%的准确率判断一个人是否同时患有其中一种以上的疾病。

美国有超过600万人受痴呆症影响，该病也是全球第七大死因。“这些疾病比我们最初想象的要复杂，而且重叠程度也超出我们的预期，”圣路易斯华盛顿大学的人类基因组学家卡洛斯·克鲁查加表示，“为了真正应对并理解其中任何一种疾病的生物学机制，我们需要将这些疾病作为一个整体来研究。”

他指出，即使症状相似，不同类型的痴呆症也需要不同的护理方案。了解这些疾病的组合，有助于制定更具针对性的精准治疗。

去年，美国食品药品监督管理局批准了首个阿尔茨海默病血液检测。市面上还有不少未经FDA批准的阿尔茨海默病检测产品。但这些检测只能检测阿尔茨海默病，无法识别其他疾病。

而新的检测可以做到。例如，它可能显示一个人同时患有阿尔茨海默病和帕金森病，并提示在检测到的异常蛋白质中，75%指向阿尔茨海默病，而较少的一部分（比如20%）指向帕金森病。凭借这一更全面的图像，克鲁查加说：“我们将能够提供更好的治疗。”

该检测基于圣路易斯华盛顿大学两家专注于阿尔茨海默病和帕金森病的诊所中3000多名患者的血液样本和医疗记录。借助一种旨在从大型数据集中发现模式的人工智能算法，研究团队将123种蛋白质的候选面板缩减至最终15种，这些蛋白质最能指示大脑中是否存在某种类型的痴呆症或帕金森病，以及其比例。这份列表包括与不同痴呆症相关的蛋白质，例如p-tau217——这是现有阿尔茨海默病血液检测的核心指标。

随后，研究人员在亚利桑那州班纳太阳健康研究所的另一组患者中验证了人工智能模型的预测结果，将血液检测结果与实际尸检捐赠的大脑组织中的发现进行了比较。

“这项检测摆脱了简单的‘是/否’阿尔茨海默病框架，而是试图反映我们在临床上经常看到的生物学复杂性，”波士顿塔夫茨医学中心的神经心理学家达维德·卡彭表示，他未参与这项研究。这对该领域来说是一个“重要的概念转变”。但他也指出，需要进行更大规模、更多样化的研究来验证这项检测。

“也许五六年前，还没有人相信阿尔茨海默病或神经退行性疾病的血液生物标志物是可能的，”克鲁查加说。但技术发展的步伐一直在加快。

他和同事们将他们的检测命名为GPND-AI，这是一个冗长名称的缩写：可推广的基于蛋白质的神经退行性疾病人工智能分类器。该团队正在与制药公司洽谈，计划将这项检测推进临床试验，并启动FDA审批流程。

不过，卡彭表示，像GPND-AI这样的检测并不能取代临床评估。“血液生物标志物可以提示潜在的病理变化，”他说，“但它们无法完全解释一个人日常如何运作、症状为何出现，也无法解释睡眠、抑郁、药物、血管疾病或抗逆力等因素如何影响认知功能。”

英文来源：

A blood test for dementia may tell you if you have more than one type
The experimental test could help doctors spot overlapping brain diseases
When something goes wrong in the brain of people with dementia, often it’s more than one thing. But it can be hard to tease apart multiple brain diseases, especially in the early stages, or even determine if more than one disease is at play. An experimental new blood test may change that.
The test measures the levels of 15 proteins in the blood to help diagnose four major neurodegenerative diseases — Alzheimer’s, Parkinson’s, frontotemporal dementia and dementia with Lewy bodies. And it can determine if a person has more than one of those diseases with 92.3 percent accuracy, researchers report in the May Alzheimer’s & Dementia.
Dementia affects more than 6 million people in the United States and is the seventh leading cause of death worldwide. “These diseases are more complex than we initially thought, and there is more overlap than we thought,” says Carlos Cruchaga, a human genomicist at Washington University in St. Louis. “In order to really address and understand the biology of the disease of any of these, we need to study all of these diseases together.”
Different dementias require different kinds of care, he says, even if the symptoms seem similar. Knowing the combination of diseases can help point toward more tailored precision treatment.
Last year the U.S. Food and Drug Administration approved the first blood test for Alzheimer’s disease. A number of other Alzheimer’s tests that do not have FDA backing are on the market. But those tests can’t detect anything more than Alzheimer’s.
The new test can. It could show that an individual has, for instance, both Alzheimer’s and Parkinson’s, and hint that of the rogue proteins detected, 75 percent signal Alzheimer’s, while a lesser amount, say, 20 percent, signal Parkinson’s. With this fuller picture, Cruchaga says, “we are going to be able to provide much better therapy.”
The test is built on blood samples and medical records from more than 3,000 patients at two clinics at Washington University in St. Louis focused on Alzheimer’s and Parkinson’s. With the help of an AI algorithm designed to find patterns in large datasets, the team narrowed a panel of 123 proteins to the final 15 that best signal whether a type of dementia or Parkinson’s might be present in the brain, and in what proportion. That list includes proteins implicated in different dementias, such as p-tau217, the cornerstone of existing Alzheimer’s blood tests.
The researchers then verified the AI model’s predictions against a separate group of patients at the Banner Sun Health Research Institute in Arizona, comparing blood test results against what was actually found in brain tissue donated at autopsy.
The test “moves away from a simple yes/no Alzheimer’s framework and instead tries to reflect the biological complexity we often see clinically,” says Davide Cappon, a neuropsychologist at Tufts Medical Center in Boston who was not involved in the study. That’s “an important conceptual shift” for the field. But he says larger, more diverse studies are needed to validate the test.
“Maybe five, six years ago, nobody believed that blood biomarkers for Alzheimer’s or neurodegenerative diseases were even possible,” Cruchaga says. But the pace of technology has been accelerating.
He and his colleagues dubbed their test GPND-AI, short for a mouthful of a name: the generalizable protein-based neurodegenerative disease artificial intelligence classifier. The team is talking with pharmaceutical companies to take the test into clinical trials and begin the FDA approval process.
Tests like GPND-AI, Cappon says, don’t replace a clinical evaluation, though. “Blood biomarkers can suggest underlying pathology,” he says. “They cannot fully explain how a person is functioning day-to-day, why symptoms started, or how factors like sleep, depression, medications, vascular disease, or resilience are influencing cognition.”