Version: 1.0.0 | Published: 24 Sep 2026 | Updated: 0 days ago
Documentation
Description:
Sixteen individuals (n = 8 early-onset Alzheimer’s disease [EOAD] cases and n = 8 cognitively normal controls) participated in the single-cell RNA sequencing (scRNA-seq) study. Study participants were recruited as part of ongoing studies of Alzheimer’s disease (AD) and related neurodegenerative diseases at the University of California, San Francisco (UCSF) Memory and Aging Center (MAC). Human PBMCs were obtained from study participants and prepared for scRNA-seq using the Chromium Single Cell 3’ v3 kit according to the manufacturer’s instructions (10x Genomics). Samples were processed in two separate batches of eight samples each, with four EOAD cases and four controls included in each batch. After sample thawing, counting, and dilution, PBMCs underwent standard 10x processing, 3’ gene expression library construction steps, and next-generation sequencing at the UCSF Genomics CoLab and Institute for Human Genetics (IHG).
Sequencing Data Processing
For each of the two batches, single-cell 3’ libraries generated from eight samples were pooled and sequenced on one lane of a NovaSeq S4 flow cell. Raw sequencing reads were aligned to GRCh38-2020-A, and feature-barcode matrices were generated using Cell Ranger version 7.1.0 with intronic reads excluded.
Quality Control
We obtained a total of 7.4 x 10^9 reads and detected ~260,000 cells across the two independent 10x and sequencing batches, yielding a moderate sequencing depth of ~30,000 mean reads/cell. We detected ~5,300 median UMI counts/cell and ~1,600 median genes/cell. There were no significant differences in the number of cells captured per sample, the number of reads per sample, or the mean read depth per sample when comparing the EOAD group to the control group. Subsequent quality-control (QC) and downstream analysis steps were performed using Seurat v4.3.0.1. QC filtering was applied to individual-sample feature-barcode matrices. After stringent QC filtering, ~182,000 cells remained for downstream analysis.
Clustering
After QC, we performed the following additional processing steps: (i) we applied sctransform v2 regularization at the individual-sample level, including mitochondrial mapping percentage as a covariate, to minimize variability due to differences in sequencing depth between samples; (ii) the 16 individual samples were integrated with FindIntegrationAnchors and IntegrateData, specifying ‘sctransform’ as the normalization method and canonical correlation analysis (CCA) as the reduction. Subsequently, PCA was performed followed by uniform manifold approximation and projection (UMAP) reduction using the first 30 PCs; clustering was performed using a resolution parameter of 0.5. This resulted in the generation of 19 clusters that were annotated via multimodal reference mapping to a large, well-characterized human PBMC dataset.
Please refer to the methods section of the supplemental online content file (see link to preprint in DOI section below) for a complete description of the methods used in this study.
Coverage
Spatial:
US
Follow Up:
Unknown
Provenance
Temporal
Accrual Periodicity:
Static
Start Date:
01 January 2026
Time Lag:
Not applicable
Accessibility
Access
Access Rights:
See dataset for details. Access data at:
Usage
Resource Creators:
Daniel W. Sirkis, Caroline Warly Solsberg, Jennifer S. Yokoyama
Observations
Statistical Population
Population Description
Population Size
Measured Property
Observation Date
Persons
16
Count
01 January 2026