AbstractBackgroundTranscriptional risk scores (TRS), a summation of polarized expression levels of functional genes, have been proposed for disease risk assessment using transcriptome profiles. Our previously developed blood‐based TRS showed associations with diagnosis and neuroimaging biomarkers for Alzheimer’s disease (AD). Here, we developed brain‐based TRS models and calculated TRS using brain tissue‐based transcriptome profiles.MethodWe integrated AD genome‐wide association study summary and expression quantitative trait locus data to prioritize target genes using Mendelian randomization and then calculated individual TRS by summing the polarized z‐scores over the target genes. We used brain transcriptome data from the Religious Orders Study and Rush Memory and Aging Project cohort (ROSMAP; N = 634) as a discovery set and the Mount Sinai Brain Bank cohort (MSBB; N = 244) as a replication set. We performed association analysis of TRS with AD diagnosis, amyloidopathy, tauopathy, and cross‐sectional and longitudinal cognitive function. We compared AD classification performance of TRS with polygenic risk scores (PRS) using a machine learning approach (Fig. 1).ResultHigher TRS values were significantly associated with AD compared to cognitively normal controls (Table 1). Higher TRS values were also related with amyloidopathy, tauopathy, poor cognitive performance, and faster cognitive decline. The AD classification performance of PRS was increased with the inclusion of TRS up to 12% with the area under the curve value of 0.842 (Fig. 2). These findings were replicated in the replication dataset.ConclusionOur results showed that brain‐based TRS is associated with AD pathology and clinical course. Furthermore, TRS improves the AD classification performance of PRS. Along with static PRS, dynamic TRS based on transcriptional expression levels at a given time could be a potential disease‐stage specific AD biomarker.