Abstract Alzheimer’s Disease (AD) exhibits a complex molecular and phenotypic profile. Investigating gene expression plays a crucial role in unraveling the disease’s etiology and progression. Transcriptome data across ethnic groups lack, negatively impacting equity in intervention and research. We employed 565 brains across six U.S. brain banks ( n= 399 non-Hispanic Whites, n =113 Hispanics, n= 12 African Americans) to generated bulk RNA sequencing from prefrontal cortex. We sought to identify cross-ancestry and ancestry-specific differentially expressed genes (DEG) across Braak stages, adjusting for sex, age at death, and RNA quality metrics. We further validated our findings using the Religious Orders Study/Memory Aging Project brains (ROS/MAP; n= 1,095) and performed metanalysis ( n= 1,660). We conducted Gene Set and Variation and Enrichment analysis (GSVA; GSEA). We employed a machine-learning approach for phenotype prediction and gene prioritization to construct a polytranscriptomics risk score (PTRS) splitting our sample into training and testing sub-samples, either randomly or by ethnicity (“ancestry-agnostic” and “ancestry-aware”, respectively). Lastly, we validated top DEG using single-nucleus RNA sequencing (snRNAseq) data. We identified several DEG associated with Braak staging: AD-known genes VGF ( P adj =3.78E- 07) and ADAMTS2 ( P ad j =1.21E-04) were consistently differentially expressed across statistical models, ethnicities, and replicated in ROS/MAP. Genes from the heat shock protein ( HSP ) family, e.g. HSPB7 ( P adj =3.78E-07), were the top differentially expressed genes and replicated in ROS/MAP. Ethnic-stratified analyses prioritized TNFSF14 and SPOCD1 as top Hispanics DEG. GSEA highlighted “ Alzheimer disease ” ( P adj =4.24E-06) and “ TYROBP causal network in microglia ” ( P adj =1.68E-08) pathways. Up- and down-regulated genes were enriched in several pathways (e.g. “ Immune response activation signal pathways” , “ Vesicle-mediated transport in synapse ”, “ cognition ”). Ancestry-agnostic and ancestry-aware PTRS effectively classified brains (AUC=0.77 and 0.73 respectively) and replicated in ROS/MAP. snRNAseq validated prioritized genes, including VGF ( downregulated in neurons; P adj =1.1 E-07). This is the largest diverse AD transcriptome in post-mortem brain tissue, to our knowledge. We identified perturbated genes, pathways and network expressions in AD brains resulting in cross- ethnic and ethnic-specific findings, ultimately highlighting the diversity within AD pathogenesis. The latter underscores the need for an integrative and personalized approach in AD studies.