Skip to content

mecfs_bio.build_system.task.fast_ct_ldsc.heritability_task

Per-trait LDSC heritability and diagnostics for the traits of a fast CT-LDSC study.

Each trait's heritability is the same own-kept-set fit the all-pairs task uses (estimate_trait_context), so the two tables agree. Columns: the alias p; observed-scale h2_obs and its delete-block jackknife SE; for binary traits the liability-scale h2_liab and SE (the observed-to-liability conversion is linear in h2_obs, so both are scaled by liability_h2(1.0, info)), null for quantitative traits; the free LDSC Intercept and its SE; lambda_gc (median kept chi-square / 0.4549); mean_chi2; n_snps (kept SNPs); n_bar (mean kept N); and the alignment diagnostics coverage, allele_mismatch_fraction and position_concordance.

Classes:

Functions:

FastCtLdscHeritabilityTask

Bases: Task

One row of heritability and alignment diagnostics per aligned trait.

Methods:

Attributes:

aligned_tasks instance-attribute 

aligned_tasks: tuple[FastCtLdscAlignedTraitTask, ...]

deps property 

deps: list[Task]

index_task instance-attribute 

index_task: FastCtLdscVariantIndexTask

meta instance-attribute 

meta: Meta

n_blocks class-attribute instance-attribute 

n_blocks: int = DEFAULT_N_BLOCKS

create classmethod 

create(
    asset_id: str,
    *,
    index_task: FastCtLdscVariantIndexTask,
    aligned_tasks: Sequence[FastCtLdscAlignedTraitTask],
    n_blocks: int = DEFAULT_N_BLOCKS,
) -> Self
Source code in mecfs_bio/build_system/task/fast_ct_ldsc/heritability_task.py 
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
@classmethod
def create(
    cls,
    asset_id: str,
    *,
    index_task: FastCtLdscVariantIndexTask,
    aligned_tasks: Sequence[FastCtLdscAlignedTraitTask],
    n_blocks: int = DEFAULT_N_BLOCKS,
) -> Self:
    return cls(
        meta=multi_trait_table_meta(asset_id),
        index_task=index_task,
        aligned_tasks=tuple(aligned_tasks),
        n_blocks=n_blocks,
    )

execute 

execute(scratch_dir: Path, fetch: Fetch, wf: WF) -> Asset
Source code in mecfs_bio/build_system/task/fast_ct_ldsc/heritability_task.py 
78
79
80
81
82
83
def execute(self, scratch_dir: Path, fetch: Fetch, wf: WF) -> Asset:
    matrix = load_trait_matrix(self.index_task, self.aligned_tasks, fetch)
    contexts = estimate_trait_contexts(matrix, n_blocks=self.n_blocks)
    out_path = scratch_dir / "fast_ct_ldsc_heritability.parquet"
    heritability_table(matrix, contexts).write_parquet(out_path)
    return FileAsset(out_path)

heritability_table 

heritability_table(
    matrix: TraitMatrix,
    contexts: Sequence[TraitLdscContext],
) -> pl.DataFrame
Source code in mecfs_bio/build_system/task/fast_ct_ldsc/heritability_task.py 
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
def heritability_table(
    matrix: TraitMatrix, contexts: Sequence[TraitLdscContext]
) -> pl.DataFrame:
    rows = []
    for alias, ctx, info, diag in zip(
        matrix.aliases, contexts, matrix.sample_infos, matrix.diagnostics
    ):
        factor = liability_h2(1.0, info)
        rows.append(
            {
                FAST_CT_LDSC_TRAIT_COL: alias,
                FAST_CT_LDSC_H2_OBS_COL: ctx.h2,
                FAST_CT_LDSC_H2_OBS_SE_COL: ctx.h2_se,
                FAST_CT_LDSC_H2_LIAB_COL: None if factor is None else factor * ctx.h2,
                FAST_CT_LDSC_H2_LIAB_SE_COL: None
                if factor is None
                else factor * ctx.h2_se,
                FAST_CT_LDSC_INTERCEPT_COL: ctx.intercept,
                FAST_CT_LDSC_INTERCEPT_SE_COL: ctx.intercept_se,
                FAST_CT_LDSC_LAMBDA_GC_COL: ctx.lambda_gc,
                FAST_CT_LDSC_MEAN_CHI2_COL: ctx.mean_chi2,
                FAST_CT_LDSC_N_SNPS_COL: ctx.n_snps,
                FAST_CT_LDSC_N_BAR_COL: ctx.n_bar,
                FAST_CT_LDSC_COVERAGE_COL: diag.coverage,
                FAST_CT_LDSC_ALLELE_MISMATCH_FRACTION_COL: diag.allele_mismatch_fraction,
                FAST_CT_LDSC_POSITION_CONCORDANCE_COL: diag.position_concordance,
            }
        )
    schema_overrides = {
        FAST_CT_LDSC_H2_LIAB_COL: pl.Float64,
        FAST_CT_LDSC_H2_LIAB_SE_COL: pl.Float64,
        FAST_CT_LDSC_POSITION_CONCORDANCE_COL: pl.Float64,
    }
    return pl.DataFrame(rows, schema_overrides=schema_overrides)