Over the past decade, scientists have accelerated efforts to better understand Alzheimer’s
disease (AD). Much progress has been made in revealing the genetic architecture of AD and its
common antecedent, mild cognitive impairment (MCI). Yet, some people who incur excessive
AD risk remain cognitively normal. Identifying risk factors for cognitive deterioration in dementia
can guide novel investigations into mechanisms underlying resilience to AD. The best-available
polygenic risk score for AD explains 1.7% of overall liability independent from the leading risk
gene, APOE (accounts for 17.4% of the variance in AD), indicating that a massive portion of
genetic liability remains unresolved. Genetic risk for cardiovascular disease contributes
additional risk for AD, thus a systems-level investigation into how cardiovascular dysfunction
interacts with neurobiological mechanisms of cognitive decline is warranted. Toward this end,
we developed a transcriptome-imputation method—the Brain Gene Expression and Network
Imputation Engine (BrainGENIE)—to measure the brain transcriptome in living individuals using
blood-based gene-expression profiles. BrainGENIE is fundamentally different from other
transcriptome-imputation methods, and captures a much larger proportion of the variance in the
brain transcriptome. BrainGENIE can predict 9–57% of the brain transcriptome, yielding an
approximate 1.8-fold increase in coverage relative to the prior “gold standard” method
PrediXcan, and which greatly improves our statistical power to detect genes and pathways
associated with disease. We have also generalized our BrainGENIE framework to impute
cardiac-specific transcriptome profiles (HeartGENIE), thereby allowing us to investigate brain-
and cardiac-specific transcriptome signatures associated with cognitive deterioration in
dementia. Our proposal contains three Specific Aims to improve our transcriptome-imputation
methods, reveal gene networks and biological pathways in brain and cardiac tissue underlying
cognitive impairment in dementia, and accurately predict an individual’s longitudinal cognitive
decline pave the way to precisely define individuals who are at risk for or resilient to AD. Aim 1:
Optimize our BrainGENIE and HeartGENIE algorithms to improve the accuracy of predicted
gene-expression levels for transcripts in the brain and cardiac tissue that are not currently well
predicted. Aim 2: Identify transcriptomic signatures of cognitive impairment in dementia with
BrainGENIE and HeartGENIE. Aim 3: Develop an neural network to accurately predict cognitive
decline longitudinally. This project will identify reveal multivariate risk factors potentially driving
cognitive decline, a critical step toward improving diagnosis, intervention, and prevention of AD.
