Output file format

Output File Documentation

This page describes the three main output files generated by PLEIO.

Summary

File Extension	Purpose	Key Columns / Contents
.txt.gz	SNP-level association results	tausq, pleio_stat, LS_stat, pleio_p, LS_p
.blup.gz	BLUP-estimated SNP effects & SEs per trait	<TraitName>, <TraitName>_se
.isf	Importance sampling–based null distribution	Empirical CDF for pleio_stat

💡 Notes

• The .isf file is required for .txt.gz p-value calculation unless using analytic alternatives.
• The .blup.gz file is optional and only generated when --blup is set.
• All outputs use SNP IDs from the --metain file as row indices.

---

Detailed Output Descriptions

1. .txt.gz – PLEIO Association Results

SNP-level association results from the PLEIO multi-trait association test.
Format: Tab-delimited, gzip-compressed.

Columns

Column	Description
SNP	Variant identifier (index column). Typically in rsID format.
tausq	Estimated variance component (τ²) from the PLEIO variance component model for this SNP. Represents the estimated genetic effect variance across traits.
pleio_stat	PLEIO test statistic computed using the variance component optimization (vcm_optimization) with trait covariance structure accounted for.
LS_stat	Least Squares (LS) test statistic from the alternative LS-based method (LS function) using observed effect sizes and standard errors.
pleio_p	P-value for the PLEIO statistic, computed using the empirical null distribution from importance sampling (.isf file).
LS_p	P-value for the LS statistic, computed analytically using the LS null distribution.

Note: SNP ID is taken from the column specified by --snp in the input

---

2. .blup.gz – Best Linear Unbiased Predictors (Random Effects)

Generated if --blup is set. Contains SNP-specific BLUP estimates of trait effects, using τ² from the .txt.gz output.
Format: Tab-delimited, gzip-compressed.

Column Naming Convention

• <TraitName> — BLUP estimate of the SNP’s effect size for that trait.
• <TraitName>_se — Standard error of the BLUP estimate.

Computation Details

• BLUP estimate:
  û = V · R^{-1} · y
• Definitions:
  • V = (R^{-1} + G^{-1})^{-1}
  • G = τ² · (genetic covariance matrix)
  • R = diag(SE) · (environmental correlation matrix) · diag(SE)
• Standard errors:
  SE(û) = sqrt(diag(V))

---

3. .isf – Importance Sampling Function (Empirical Null Distribution)

Stores the empirical cumulative distribution of the PLEIO statistic under the null hypothesis, estimated by importance sampling.

Format: Tab-delimited text.

Purpose

• Used to convert pleio_stat values into empirical p-values (pleio_p) in .txt.gz.
• Enables accurate tail probability estimation without extreme permutation counts.

---