KASM

The Kernel Allelic Series Model (KASM) integrates functional annotations to form a covariance kernel describing how allelic effects within a gene covary. Using data from exome association studies, KASM estimates the heritability of rare protein coding variants, its enrichment in structural and functional annotations, and the concordance of effects between pairs of variants. Our paper describes the methodology and applications in detail.

KASM can be installed with pip install kasm-pkg or uv add kasm-pkg.

Documentation · Quick start · Development

Quick start

Here is an example of using KASM to estimate heritability from exome-wide association statistics. KASM takes variant-level association statistics: one row per variant, with a canonical chr:pos:ref:alt ID, its gene, the effect estimate and SE, allele frequency, and annotations used as burden weights. Within-gene LD comes from an LD matrix file passed to Dataset as ldm; without one, as in this toy example, KASM assumes no LD between variants.

import kasm
import polars as pl

variant_df = pl.DataFrame(
    {
        "variant": [f"1:{pos}:A:G" for pos in range(100, 1000, 100)],
        "gene": ["G1"] * 3 + ["G2"] * 3 + ["G3"] * 3,
        "beta": [-0.12, 0.19, -0.15, 0.01, 0.17, -0.05, 0.15, 0.25, 0.22],
        "se": [0.10, 0.12, 0.09, 0.11, 0.10, 0.13, 0.08, 0.10, 0.12],
        "af": [2e-4, 5e-5, 3e-4, 1e-4, 2e-4, 4e-5, 5e-4, 2e-4, 6e-5],
        "constant": 1.0,
        "score1": [0.9, 0.1, 0.0, 0.9, 0.1, 0.0, 0.1, 0.8, 0.9],
        "score2": [0.8, 0.3, 0.1, 0.7, 0.2, 0.0, 0.3, 0.6, 0.9],
    }
)
print(variant_df)

dataset = kasm.Dataset(variant_df, beta_col="beta", se_col="se", sample_size=300_000)
model = kasm.Model(
    dataset,
    kasm.KernelOptions(
        maf_col="af",
        burden_cols=["constant", "score1", "score2"],
        overdispersion_cols=["constant"],
    ),
).compute_suffstats()
fit = kasm.fit_random_effects(model, exclude_genes=[], n_jk_blocks=3)
print(f"heritability: {fit.hsq:.2e} (SE {fit.hsq_se:.2e})")
print(fit.components.select("kernel", "kernel_coef", "kernel_coef_se"))

See more examples in the documentation.

Development

Development uses uv:

uv run --extra dev pytest
uv run --extra dev pytest --full
uv run --extra dev pre-commit install