Python API Cookbook

The CLI is the main path for routine runs. Use the Python API when you want to inspect objects, run one step in a notebook, or connect CREDTOOLS to your own workflow.

Load a Locus Set

import pandas as pd
from credtools.locus import load_locus_set

loci_info = pd.read_csv("loci_list.txt", sep="\t")
locus_info = loci_info[loci_info["locus_id"] == "locus_1"]

locus_set = load_locus_set(locus_info, if_intersect=True)
print(locus_set)

load_locus_set expects each row to have prefix, popu, cohort, sample_size, chr, start, end, and locus_id.

Run QC for One Locus

from credtools.qc import locus_qc, locus_qc_summary

metrics = locus_qc(locus_set, out_dir="work/qc/locus_1")
summary = locus_qc_summary(metrics)
print(summary)

Open the detailed data frames directly:

expected_z = metrics["expected_z"]
dentist_s = metrics["dentist_s"]
compare_maf = metrics["compare_maf"]

Run Meta-Analysis on One Locus

from credtools.meta import meta

meta_locus_set = meta(locus_set, meta_method="meta_all")
print(meta_locus_set)

Common meta_method values:

Method	Use
meta_all	combine all input rows into one meta-analyzed locus
meta_by_population	combine cohorts within each population
no_meta	keep rows separate

Run Fine-Mapping

from credtools.credtools import fine_map

creds = fine_map(
    locus_set,
    tool="susie",
    max_causal=5,
    coverage=0.95,
)

print(creds.n_cs)
print(creds.pips.sort_values(ascending=False).head())

Create the same enhanced PIP table that the CLI writes:

pips = creds.create_enhanced_pips_df(locus_set)
pips.to_csv("pips.txt.gz", sep="\t", index=False, compression="gzip")

Run ABF Without LD

The CLI loader expects LD files for loci_list.txt workflows. If you want to run true no-LD ABF, construct a Locus object directly.

import pandas as pd
from credtools.locus import Locus
from credtools.wrappers.abf import run_abf

sumstats = pd.read_csv("single_locus.sumstats.gz", sep="\t")

locus = Locus(
    popu="EUR",
    cohort="UKBB",
    sample_size=400000,
    sumstats=sumstats,
    locus_start=50_000_000,
    locus_end=50_500_000,
    ld=None,
)

creds = run_abf(locus, coverage=0.95)

Plot from Python

from credtools.plot import plot_summary_qc, plot_locusplot

plot_summary_qc("work/qc/qc.txt.gz", output_file="qc_summary.png")
plot_locusplot("work/results/locus_1", output_file="locus_1.png")

For notebook work, omit output_file and display the returned Matplotlib figure.

Set an External Tool Path

The CLI uses executables on PATH. In Python, you can set a tool path directly:

from credtools.utils import tool_manager

tool_manager.set_tool_path("finemap", "/opt/finemap/finemap")

This affects the current Python process.