A singular value decomposition algorithm to identify early dysfunctional molecular pathways in Alzheimer’s disease

Abstract

AbstractBackgroundMicroarray technology has been an effective tool to capture gene expression of thousands of genes from multiple molecular pathways in human brain tissue. Many large expression datasets exist maintaining information about differences in gene expression related to Alzheimer’s disease processes. Often groups employ comparative analysis strategies on the data resulting in differential expression between disease and controls.MethodUsing available expression datasets with multiple brain regions as a function of Braak‐staging, we look for expression signatures that do not necessarily increase nor decrease with disease progression, but rather have highly variable patterns indicating a molecular process dysregulation. Here we use a variation on singular value decomposition (SVD) as a rigorous approach to this type of analysis thus providing a characterization of the structure of the data as well as information about the relationship among genes.ResultAnalysis of the sequential singular values generated determines the number of components that capture the data essential features. The magnitude of the singular values identified dysfunctional expressed transcripts at early stage in the disease before the appearance of tau protein neurofibrillary tangles. It follows that by pathway analysis we can determine upstream players in the affected biological processes pathway and design appropriate experiments to validate the early dysregulated molecular pathways.ConclusionUsing a novel approach to analyze expression data, we identified transcripts displaying dysregulated expression signatures at early stages of disease in brain regions not yet affected by neurofibrillary tangles. The relationships of these transcripts and pathways enriched then guided us to explore upstream events in the early disease cascade.