Researchers at Columbia University Mailman School of Public Health have developed a novel computational pipeline designed to identify protein biomarkers associated with complex diseases, including Alzheimer’s disease (AD). The new tool analyzes biomarkers that can induce 3D structural changes in proteins, providing insights into disease mechanisms, and highlighting potential targets for therapeutic intervention. In their reported study the team applied the tool to proteomics data from the UK Biobank, to identify seven proteins that were significantly associated with the risk of AD. The findings, they suggested, could lead to improved early detection and treatment strategies for AD, which has long eluded effective therapies.
“Alzheimer’s disease is defined by amyloid-beta plaques and tau neurofibrillary tangles in the brain, which accumulate decades before symptoms,” said Zhonghua Liu, ScD, assistant professor of biostatistics at Columbia Mailman School, and senior investigator. “Current early diagnostics are either resource-intensive or invasive. Moreover, current AD therapies targeting amyloid-beta provide some symptomatic relief and may slow disease progression but fall short of halting it entirely. Our study highlights the urgent need to identify blood-based protein biomarkers that are less invasive and more accessible for early detection of Alzheimer’s disease. Such advancements could unravel the underlying mechanisms of the disease and pave the way for more effective treatments.”
Liu and colleagues reported on their findings in Cell Genomics. In their paper, titled “Deciphering causal proteins in Alzheimer’s disease: A novel Mendelian randomization method integrated with AlphaFold3 for 3D structure prediction,” the team noted, “…our pipeline holds promising implications for drug target discovery, drug repurposing, and therapy development.”
AD is the primary cause of dementia globally, exerting a considerable strain on healthcare resources, the authors noted. “Despite extensive efforts, the etiology and pathogenesis of AD are still unclear, and strategies aimed at impeding or delaying its clinical advancement have largely remained challenging to achieve … it is imperative and urgent to identify causal protein biomarkers to elucidate the underlying mechanisms of AD and to expedite the development of effective therapeutic interventions for AD.”
The new computational pipeline developed by Liu et al., named MR-SPI (Mendelian Randomization by Selecting genetic instruments and Post-selection Inference), has several key advantages, the team noted. Unlike traditional methods, MR-SPI does not require a large number of candidate genetic instruments, or instrumental variables; IVs (e.g., protein quantitative trait loci; pQTLs) to identify disease-related proteins. Rather, MR-SPI is a powerful tool designed for studies with only a limited number of genetic markers available.
“In this paper, we develop a novel all-in-one pipeline for causal protein biomarker identification and 3D structural alteration prediction using large-scale genetics, proteomics, and phenotype/disease data …” the team explained. “Specifically, we propose a two-sample MR method and algorithm that can automatically select valid pQTL IVs and then performs robust post-selection inference (MR-SPI) for the causal effect of proteins on the health outcome of interest.”
Liu further commented, “MR-SPI is particularly valuable for elucidating causal relationships in complex diseases like Alzheimer’s, where traditional approaches struggle. The integration of MR-SPI with AlphaFold3—an advanced tool for predicting protein 3D structures—further enhances its ability to predict 3D structural changes caused by genetic mutations, providing a deeper understanding of the molecular mechanisms driving disease.”
Applying the pipeline to data from the UK Biobank, which includes 54,306 participants, and a genome-wide association study (GWAS) of Alzheimer’s disease with 455,258 subjects (71,880 AD cases and 383,378 controls), the research team identified seven key proteins—TREM2, PILRB, PILRA, EPHA1, CD33, RET, and CD55—that exhibit structural alterations linked to Alzheimer’s risk.
“We discovered that certain FDA-approved drugs already targeting these proteins could potentially be repurposed to treat Alzheimer’s,” Liu added. “Our findings underscore the potential of this pipeline to identify protein biomarkers that can serve as new therapeutic targets, as well as provide opportunities for drug repurposing in the fight against Alzheimer’s.”
The study’s findings also indicate that MR-SPI could have wide-reaching applications beyond Alzheimer’s disease, offering a powerful framework for identifying protein biomarkers across various complex diseases. The ability to predict 3D structural changes in proteins opens up new possibilities for drug discovery and the repurposing of existing treatments.
“By combining MR-SPI with AlphaFold3, we can achieve a comprehensive computational pipeline that not only identifies potential drug targets but also predicts structural changes at the molecular level,” Liu concluded. “This pipeline offers exciting implications for therapeutic development and could pave the way for more effective treatments for Alzheimer’s and other complex diseases.”
Co-author Gary W. Miller, PhD, Columbia Mailman vice dean for research strategy and innovation and professor, department of environmental health sciences, further noted, “By leveraging large cohorts with biobanks, innovative statistical and computational approaches, and AI-based tools like AlphaFold this work represents a convergence of innovation that will improve our understanding of Alzheimer’s and other complex diseases.”
The authors concluded: “Our pioneering pipeline, for the first time, integrates the identification of causal protein biomarkers for health outcomes and the subsequent analysis of their 3D structural alterations into a unified framework, leveraging increasingly publicly available GWAS summary statistics for health research.”
