Blood-Based Genetic and Gene Expression Scores Together Boost Alzheimer's Detection
Combining polygenic and transcriptional risk scores from blood raises Alzheimer's detection accuracy, offering a less invasive screening path.
Summary
Researchers tested whether combining two types of blood-based risk scores could better identify Alzheimer's disease than either score alone. Polygenic risk scores (PRS) capture inherited genetic risk, while transcriptional risk scores (TRS) reflect how genes are currently being expressed in the blood. Using data from two cohorts — one American, one Korean — the team found that people with high scores on both measures had 2.5 to 3.4 times greater odds of having Alzheimer's compared to those with low scores on both. When combined into a single model, the integrated score achieved an AUC of 0.705, outperforming PRS alone at 0.635. This approach could eventually offer a cheaper, less invasive alternative to current biomarkers like PET scans or cerebrospinal fluid tests.
Detailed Summary
Alzheimer's disease affects tens of millions worldwide, yet early detection remains a major clinical challenge. Current gold-standard biomarkers — amyloid PET imaging and cerebrospinal fluid analysis — are expensive, invasive, and inaccessible to most patients. Blood-based risk tools that integrate genetic and gene expression data could change that equation significantly.
This study examined whether combining polygenic risk scores (PRS), which aggregate inherited genetic variants associated with Alzheimer's risk, with transcriptional risk scores (TRS), which capture real-time gene expression patterns from blood, could improve early disease detection. Researchers used genome-wide genotype and blood transcriptome data from two ancestrally distinct cohorts: the Alzheimer's Disease Neuroimaging Initiative (ADNI, N=313, primarily American) and Seoul National University Bundang Hospital (SNUBH, N=173, Korean). Logistic regression and machine learning models were applied to evaluate predictive performance.
The results were encouraging. Individuals with high PRS and TRS values showed 2.5-fold greater odds of Alzheimer's in the ADNI cohort and 3.4-fold greater odds in the SNUBH cohort compared to those with low values on both scores. The integrated PRS-TRS model achieved an AUC of 0.705, meaningfully outperforming the PRS-only model at 0.635. The consistency across two ethnically distinct populations strengthens confidence in the approach's generalizability.
The clinical implications are notable. A blood-based, non-invasive screening tool that integrates static genetic risk with dynamic gene expression could enable earlier, broader population screening — particularly in settings where PET or lumbar puncture is impractical. This could be especially valuable for identifying at-risk individuals before symptom onset.
Important caveats apply. Sample sizes were modest, and an AUC of 0.705, while an improvement, still falls short of clinical-grade diagnostic accuracy. The summary is based on the abstract only, so methodological details and subgroup analyses cannot be fully evaluated. Replication in larger, more diverse cohorts is needed before clinical translation.
Key Findings
- High PRS plus high TRS associated with 2.5–3.4x greater Alzheimer's odds across two distinct populations.
- Integrated PRS-TRS model achieved AUC of 0.705, outperforming PRS alone (AUC 0.635).
- Blood-based approach avoids invasive procedures like lumbar puncture or PET imaging.
- Findings replicated across American and Korean cohorts, suggesting cross-ancestry generalizability.
- Combining static genetic and dynamic gene expression data improves Alzheimer's classification.
Methodology
The study used genome-wide genotype and blood transcriptome data from two cohorts (ADNI, N=313; SNUBH, N=173) representing distinct ancestries. Logistic regression and machine learning models were used to assess individual and combined predictive performance of PRS and TRS. Cross-cohort replication strengthens the validity of findings.
Study Limitations
Sample sizes were modest (313 and 173 participants), and an AUC of 0.705 is insufficient for standalone clinical diagnosis. The summary is based on the abstract only, limiting evaluation of methodological details, covariate adjustments, and subgroup analyses. Larger, more diverse validation cohorts are needed before clinical implementation.
Enjoyed this summary?
Get the latest longevity research delivered to your inbox every week.