The main inference function for identifying mediating genes using Mr. PEG.

Performs mediation effect estimation integrating GWAS, eQTL, and perturbation data with permutation-based null distribution inference.

Parameters:

beta: Array | ndarray | bool_ | number | bool | int | float | complex¶

GWAS effect sizes, array of shape (k,).

se: Array | ndarray | bool_ | number | bool | int | float | complex¶

GWAS standard errors, array of shape (k,).

eqtl: Array | ndarray | bool_ | number | bool | int | float | complex¶

eQTL Z-scores, array of shape (k,).

perturb: Array | ndarray | bool_ | number | bool | int | float | complex¶

Perturbation effect size matrix, array of shape (k, t), where k is the number of perturbed genes and t is the number of downstream target genes.

ld: Array | ndarray | bool_ | number | bool | int | float | complex¶

Linkage disequilibrium matrix, array of shape (k, k).

perm_number: int = 500¶

Number of permutations for null distribution estimation. Must be positive; values below 100 may yield inaccurate p-values.

seed: int = 12345¶

Random seed for permutation testing. Must be a positive integer.

alt: bool = False¶

Whether to use the alternative distribution assumption for the null model.

Returns:

Array of shape (t, 6) where each row corresponds to a downstream gene and columns are:

• gamma: Estimated mediation effect size.

• gamma_se: Standard error of the effect size.

• gamma_p: P-value from t-test.

• gamma_perm_mean: Mean of the permutation null distribution.

• gamma_perm_z: Z-score relative to permutation null.

• gamma_null_p: Two-sided p-value from permutation null.

Raises:

ValueError – If seed <= 0, perm_number <= 0, or input dimensions mismatch.

Example

>>> import numpy as np
>>> from mrpeg.peg import infer_peg
>>> beta = np.random.randn(10)
>>> se = np.ones(10) * 0.1
>>> eqtl = np.random.randn(10)
>>> perturb = np.random.randn(10, 5)
>>> ld = np.eye(10)
>>> results = infer_peg(beta, se, eqtl, perturb, ld, perm_number=10, seed=42)
>>> results.shape
(5, 6)