Skip to content

mecfs_bio.build_system.task.pipes.compute_se_pipe

Compute standard error in GWAS summary statistics.

Classes:

ComputeSEIfNeededPipe

Bases: DataProcessingPipe

Methods:

process 

process(x: LazyFrame) -> narwhals.LazyFrame
Source code in mecfs_bio/build_system/task/pipes/compute_se_pipe.py 
41
42
43
44
45
def process(self, x: narwhals.LazyFrame) -> narwhals.LazyFrame:
    cols = x.collect_schema().keys()
    if GWASLAB_SE_COL not in cols:
        return ComputeSEPipe().process(x)
    return x

ComputeSEPipe

Bases: DataProcessingPipe

Pipe to compute standard error in GWAS summary statistics using p-value and beta

Methods:

process 

process(x: LazyFrame) -> narwhals.LazyFrame
Source code in mecfs_bio/build_system/task/pipes/compute_se_pipe.py 
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
def process(self, x: narwhals.LazyFrame) -> narwhals.LazyFrame:
    cols = x.collect_schema().keys()
    assert GWASLAB_SE_COL not in cols
    assert GWASLAB_P_COL in cols
    assert GWASLAB_BETA_COL in cols
    collected: pd.DataFrame = x.collect().to_pandas()
    pvals: np.ndarray = collected[GWASLAB_P_COL].to_numpy()
    z_score: np.ndarray = abs(scipy.stats.norm.ppf(1 - pvals / 2))
    z_score_2: np.ndarray = abs(scipy.stats.norm.ppf(pvals / 2))
    min_z_score = np.minimum(
        z_score, z_score_2
    )  # this is to resolve numerical issues when (1-pval) is numerically equal to 1
    se = abs(collected[GWASLAB_BETA_COL] / min_z_score)
    collected[GWASLAB_SE_COL] = se
    return narwhals.from_native(collected).lazy()