High-performance protein signatures in the battle against Alzheimer’s Disease & dementia

Existing biomarkers for Alzheimer’s Disease (AD) are suboptimal for distinguishing between AD and non-AD dementias, as well as in characterizing the complex and highly diverse nature of AD pathology. The pioneering work conducted by researchers in Professor Charlotte Teunissen’s team at Amsterdam University Medical Center presents a remarkable illustration of how large-scale biomarker discovery can uncover informative protein signatures using scalable technology. The discovery phase encompassed approximately 800 individuals across various AD stages, non-AD dementias, and healthy controls, with their cerebrospinal fluid (CSF) proteomes characterized using multiple Olink Target 96 panels to include approximately 700 proteins covering a broad range of pathways and disease-related biologies.

This revealed nearly 100 AD-specific dysregulated CSF proteins. Through data-driven modeling, the researchers identified an 8-protein signature that effectively discriminated AD from controls, yielding an impressive area under the curve (AUC) performance of 0.96. Furthermore, they identified a 9-protein signature that effectively differentiated AD from non-AD dementias with an AUC of 0.87. Subsequent verification of these signatures in additional cohorts led to the construction of a custom Olink Focus panel comprised of 12 proteins.

Development of a custom panel without the complications of a cross-technology gap

Using the same PEA technology to transition the biomarker discovery findings into a practical lower plex custom panel circumvented any additional verification and development work that would typically be encountered by switching protein technologies when changing the scale of analysis, ensuring a smooth and efficient process. In a successful validation study in independent cohorts, the custom panel exhibited an AUC of 0.95 for AD versus controls and an AUC of 0.79 for AD versus non-AD dementia. This custom panel laid the groundwork for evaluating the potential added value of these markers in routine diagnosis and clinical trials of drugs targeting distinct pathomechanisms of AD.

Development of a protein panel for dementia with Lewy bodies (DLB)

The broad biomarker screening that preceded the custom panel enabled not only a powerful diagnostic model for AD vs healthy controls, but also provided discrimination between AD and other forms of dementia. In a recent follow-up study, a similar strategy was used to look for biomarkers in patients suffering from dementia with Lewy bodies (DLB), a relatively common form of age-related dementia that can be difficult to diagnose and manage due to significant overlap of clinical and pathological presentations with AD. The group again used multiple Olink Target 96 panels to measure >650 proteins in CSF samples from patients with DLB, AD and cognitively unimpaired controls. This identified more than 50 DLB-associated proteins, with pathway analysis showing enrichment in myelination processes. The dopamine biosynthesis enzyme DDC was the strongest dysregulated protein, and this marker alone could effectively discriminate DLB from controls (AUC=0.91), as well as DLB from AD (AUC=0.81). To improve the discriminatory performance between DLB and AD, classification modeling was used to identify a 7-protein signature (DDC, CRH, MMP-3, ABL1, MMP-10, THOP1, FCER2) with significantly better performance than DDC alone (7-protein AUC=0.93).

As in the previous Alzheimer’s study, the team then designed a custom PEA panel comprising 6 of the 7 proteins from the model to seamlessly obtain a practical tool for further validation without the need to change technologies from the discovery phase. The custom panel was then successfully validated in several independent cohorts, showing high performance (AUCs>0.90) for discrimination of DLB vs both AD and controls, demonstrating that this tool offers highly promising opportunities for testing in diagnostic and clinical trial settings.