Re-Analysis of LDSC heritability estimates in Tan et al. 2024 and Howe et al. 2022

This repository contains code to re-analyze the population and direct/sibling GWAS data from Tan et al. 2024 and Howe et al. 2022 using a modern LDSC model.

The following files are required

• tan_2024_cognitive_performance.gz: The raw Tan et al. GWAS summary data, available from the SSGAC data portal
• ldsc: LD-score regression, available from github or by running git clone https://github.com/bulik/ldsc.git
• The baselineLD_v2.2 model, available from the Alkes Group Google bucket and at this direct link.
• The base model weights, available from the Alkes Group Google bucket and at this direct link.
• The base (no annotation) model, available from the Alkes Group Google bucket and at this direct link.

The following scripts will perform the analysis

run_all.sh : Generates LDSC-format summary stats from the Tan et al. 2024 input data and runs both versions of the LDSC model run_howe.sh : Same as above but for running the Howe et al. summary stats (which are provided in this repository in LDSC format) plot.R : Generate barplots (using hard-coded numbers)

The following files are produced

*.log files from each LDSC run contain all of the heritability estimates and can be queried as follows:

$ grep "Total Observed scale h2" *.log
Howe.Cog.direct.LDSC_h2.baselineLD_model.log:Total Observed scale h2: 0.1146 (0.0674)
Howe.Cog.direct.LDSC_h2.base_model.log:Total Observed scale h2: 0.1311 (0.0434)
Howe.Cog.population.LDSC_h2.baselineLD_model.log:Total Observed scale h2: 0.1565 (0.0399)
Howe.Cog.population.LDSC_h2.base_model.log:Total Observed scale h2: 0.243 (0.0276)
tan_2024_cognitive_performance.direct.LDSC_h2.baselineLD_model.log:Total Observed scale h2: 0.1233 (0.0496)
tan_2024_cognitive_performance.direct.LDSC_h2.base_model.log:Total Observed scale h2: 0.1661 (0.0303)
tan_2024_cognitive_performance.population.LDSC_h2.baselineLD_model.log:Total Observed scale h2: 0.1261 (0.0195)
tan_2024_cognitive_performance.population.LDSC_h2.base_model.log:Total Observed scale h2: 0.1855 (0.0161)

Acknowledgements

Thank you to the authors of Tan et al. 2024 for making all data and code easily available with the pre-print. Thank you to the authors of Howe et al. 2022 for making all data available.