ROSMAP snRNA-seq pseudo-bulk gene expression QTL

Contact

Hao Sun (eQTL), Masashi Fujita (eQTL), Haochen Sun (fine-mapping), Jiajun Tao (replication)

Study Overview

• Sample information: ROSMAP/ROSMAP_Pseudo_Bulk_sample_attributes.csv.
• Lab protocol: ROSMAP/ROSMAP_Pseudo_Bulk_lab_protocol.csv.
• Computational protocol: ROSMAP/ROSMAP_Pseudo_Bulk_computational_protocol.csv.
• QTL summary statistics output: ####/####.qtl_results.csv.
• Fine-mapping results individual level data model: ####/####.susie.csv.
• Fine-mapping results summary statistics model: ####/####.susie_rss.csv.

Analysis Status

TransQTL association: Finished.

Dataset Description

Path(s) to genotype matrix

Using MatrixQTL pipeline (by Masashi)

1. genotype is an all-chromosome, all-samples vcf collection
2. The original gz vcf is gzipped but not bgzipped, thus cannot tabix -p
3. The vcf is not imputed.

• Dosage file. The number of ALT allele were counted per donor.

• SNP position file in GRCh38

• VCF file used to generate above files. This is a subset of ROSMAP WGS VCF.

• The original VCF files of ROS/MAP WGS is here (N = 1,196; GRCh37):

• A summary of quality control is here:

• Liftover of the above VCFs from GRCh37 to GRCh38.

• Sorted positions of SNPs, added rsID in dbSNP154, and renamed chromosomes (e.g. 1 to chr1).

• 424 donors extracted for snRNAseq and applied filtering of MAF, HWE, etc.

Path(s) to covariate data matrix

Using MatrixQTL pipeline (by Masashi)

Here, I use astrocytes as an example. But all other cell types have the same folder structure.

Covariates of eQTL analysis are sex, age, PMI, study, total genes detected, top 3 genotype PCs, and up to 30 expression PCs.

Using TenorQTL pipeline (by Hao)

Path(s) to QTL results

Using MatrixQTL pipeline (by Masashi)

Take astrocytes as an example,

df <- readRDS("matrix-eqtl.rds")$cis$eqtl

Using TenorQTL pipeline (by Hao)

Association scan using TensorQTL and summary statistics standardization

• TensorQTL.ipynb provides the pipeline to generate TensorQTL cis association results for all QTLs.
• ROSMAP CUIMC1 snuc eQTL provides information about the input files for TensorQTL cis association in the base_params variable in [generate_command_1].
• ROSMAP MIT snuc eQTL provides information about the input files for TensorQTL cis association in the base_params variable in [generate_command_1].
• ROSMAP_mega_eQTL provides information about the input files for TensorQTL cis association in the base_params variable in [generate_command_1].

Path(s) to cis-QTL association testing

output of TensorQTL.ipynb

Path(s) to fine-mapping with SuSiE RSS model

Fine-mapping model objects (SuSiE-RSS, .rds) are available in the finemapping models folder (syn69670592).

Path(s) to TWAS models

Pre-trained TWAS weight models for this context: syn69670600

Quantile TWAS (qTWAS) models: syn69670611

Path(s) to colocalization with SuSiE-coloc / ColocBoost

Multi-context colocalization models (ColocBoost): syn69670597

AD GWAS–xQTL colocalization results: syn69865816

AD GWAS–xQTL colocalization models: syn69670630

Links to QTL analysis notebooks

CUIMC1: Preprocess_bundle provides commands to preprocess genotype, phenotype and covariate data all at once. Phenotype_preprocessing shows the commands used for the phenotype data processing and preparation steps for all cell types. Cell-specific phenotype preprocessing are listed here in different folders.

MIT: Preprocess_bundle provides commands to preprocess genotype, phenotype and covariate data all at once. Genotype_pca provides steps for PCA analysis for genotype data. Phenotype_preprocessing shows the commands used for the phenotype data processing and preparation steps. Covariates_preprocessing shows the commands used for the covariate data processing and preparation steps.

pseudo_bulk_eQTL_mega: Genotype_pca provides steps for PCA analysis for genotype data. Phenotype_preprocessing shows the commands used for the phenotype data processing and preparation steps. Covariates_preprocessing shows the commands used for the covariate data processing and preparation steps.