An information‐sharing framework of patient label propagation method to reveal distinct blood biomarkers and signatures in Alzheimer’s disease subgroups
Xiaoqing Huang, Nur Jury Garfe, Cristian A Lasagna Reeves, Kun Huang, Jie Zhang- Psychiatry and Mental health
- Cellular and Molecular Neuroscience
- Geriatrics and Gerontology
- Neurology (clinical)
- Developmental Neuroscience
- Health Policy
- Epidemiology
Abstract
Background
Alzheimer’s Disease (AD) is a complex and heterogenous dementia and the early detection remains very challenging. Although progresses have been made to identify serum biomarkers associated with AD, it is still difficult to generalize the findings into clinical setting and trace back to learn how these AD biomarkers are progressed through the regulatory pathways and how they play the roles in the disease mechanisms. Furthermore, distinct patterns or signatures among atypical progressing subgroups has not been investigated.
Method
We defined four subgroups of AD patients based on the trajectories of tauopathy and cognitive decline. Using the blood transcriptomic data from ROSMAP and ADNI, we performed hierarchical clustering for the top 1000 variable genes to obtain the gene signatures in each group. An NMF‐based decomposition and network assisted label propagation method was applied. Finally, we use the inferred label for ADNI patients to assess the distinct progression patterns for each AD subtype and validate and uncovering the genes and their enriched pathways that are associated with flagged progression biomarkers.
Result
Our preliminary results show that two gene expression clusters exist for both ROSMAP and ADNI data. Our previous study identified the activation of calcium channel related pathway exhibits a detrimental role in AD, with representative genes such as TGFB1, and S100A12. We observe the same pattern for these two genes in a subset of patients from ADNI gene expression data with an enrichment in neovascularization processes. We are now propagating patient group labels with the prior evidence of the alignment between these two data sets and investigating the primary genes linked to specific biomarker progression patterns through longitudinal analyses.
Conclusion
We identified several altered pathways and genes that potentially render atypical AD progression patterns reflected by various biomarkers. Future work is undergoing to further explore the link between the gene signatures in each subgroup and the blood transcriptomic sub‐clusters, as well as the comparisons of the two blood transcriptomic data for potential longitudinal clinical information mapping between the two datasets.