Polygenic scoring for each local ancestry background#

[1]:

import pandas as pd
import numpy as np
import admix
import os

[2]:

# load the example dataset
dset = admix.io.read_dataset("example_data/CEU-YRI")

[3]:

# construct an example SNP weights dataframe (WEIGHT column contains the per-allele effect sizes)
df_weights = dset.snp[["CHROM", "POS", "REF", "ALT"]].copy()
df_weights["WEIGHT"] = np.random.normal(size=len(df_weights))
df_weights

[3]:

	CHROM	POS	REF	ALT	WEIGHT
snp
22:16406147:A:G	22	16406147	A	G	-0.395390
22:16551808:T:C	22	16551808	T	C	1.332264
22:16573830:T:C	22	16573830	T	C	0.062149
22:16575525:T:C	22	16575525	T	C	0.428249
22:16576248:G:T	22	16576248	G	T	0.263856
...	...	...	...	...	...
22:50739662:G:A	22	50739662	G	A	-0.963074
22:50743331:A:G	22	50743331	A	G	-0.816075
22:50772964:T:C	22	50772964	T	C	-0.379005
22:50774447:A:C	22	50774447	A	C	0.983688
22:50780578:G:A	22	50780578	G	A	0.721975

15357 rows × 5 columns

[4]:

# calculate PGS for each local ancestry background
pgs_df = admix.data.calc_partial_pgs(dset=dset[:, 0:100], df_weights=df_weights)

admix.data.calc_partial_pgs: 100%|██████████████████████████████████████████████████████████████████████████████████████████████████| 100/100 [00:00<00:00, 1268.91it/s]

[5]:

# per-allele effects x genotype dosage within each local ancestry background
pgs_df

[5]:

	ANC1	ANC2
indiv
Sample_1	35.638558	118.706099
Sample_2	-30.856414	60.780757
Sample_3	61.098364	-46.071822
Sample_4	-27.939297	28.420301
Sample_5	-2.792197	213.910173
...	...	...
Sample_96	0.000000	45.962131
Sample_97	-58.187541	24.903466
Sample_98	-30.823394	39.505350
Sample_99	0.000000	84.651254
Sample_100	-46.359911	90.501891

100 rows × 2 columns

Alternatively, this can be calculated in a command line:

admix calc-partial-pgs \
    --plink-path <plink2_prefix>.pgen \
    --weights-path <weight_tsv_path> \
    --out out