FunGen-xQTL Resources

LD panelReferenceADSP

ADSP LD Sketch Panel

A compressed LD reference built from the same ADSP Release 4 whole genomes as the in-sample LD reference panel (16,571 European-ancestry individuals). The genotype matrix is multiplied by a random matrix with B = 10,000 columns, so the panel holds 10,000 “pseudo-people” instead of real individuals. LD computed from the sketch matches in-sample LD on average, and the panel is far smaller than the genotypes it comes from.
Cohort
ADSP
Modality
LD panelReference
Sample size
16,571 samples
Access
ControlledData use certificate
Lead analysts
Jenny Anak Empawi

Get the data

Controlled access. Request data access for ADSP on the AD Knowledge Portal; you need a free account there. See How to get the data.

WhereIDAccessWhat you get
AD Knowledge Portalsyn75082260ControlledFiles described below
shell
pip install synapseclient
synapse login
synapse get -r syn75082260

When to use it

Use it as the LD reference for summary-statistics fine-mapping (SuSiE-RSS) and TWAS when the in-sample panel cannot be shared or is too large to move. The FGMB Atlas uses it for RWAS association testing and fine-mapping.

Privacy

No row of the sketch is one person’s genotypes. As long as the random projection matrix and its seed are not released, a sketch reveals no more about who is in the cohort than the exact cohort LD would. This compares two ways of sharing LD; it is not an absolute privacy guarantee such as differential privacy (Sun et al., manuscript).

Format

PLINK 2 genotype files (.pgen, .pvar, .psam) with 10,000 pseudo-samples, in the format made by the protocol’s RSS LD sketch pipeline.

Data Availability

ResourceAD Knowledge Portal ID
ADSP LD sketch (PLINK 2)syn75082260
  1. Building the LD sketch: https://github.com/StatFunGen/xqtl-protocol/blob/main/code/SoS/reference_data/rss_ld_sketch.ipynb
  2. Fine-mapping with summary statistics: https://github.com/StatFunGen/xqtl-protocol/blob/main/code/SoS/mnm_analysis/mnm_methods/rss_analysis.ipynb

Reference

Sun, Cao, Zou et al. Robust fine-mapping of molecular and complex traits with shareable LD references. Manuscript.