How to remove duplicated SNPs based on MAF with Plink?

Question

[giulia.trauzzi]

Hi everyone,

I am trying to remove duplicated SNPs from my pgen dataset. These duplicated SNPs are the result of splitting multiallelic loci but now I just want to retain only the genotype that has higher maf, the most common. Is there a way to do this with Plink2? Considering that the most common genotype is not always the first instance in the data so I cannot use the --rm-dup first etc... Is there a way I can do this?

Many thanks.

Giulia

[hakimbazol]

Since the MAF values are different, is it possible to filter your result based on MAF first? And then see whether the duplicates are still there or not.

[Semir]

# Calculate allele frequencies for all variants
plink2 --pfile ${input_prefix} \
       --freq \
       --out ${params.output_prefix}

Python script to process the frequency file:

import pandas as pd
# Read the frequency file
df = pd.read_csv("${freq_file}", delim_whitespace=True)

# Extract the base variant ID (removing the split variant suffix if present)
df['base_id'] = df['ID'].str.split('_').str[0]

# For each group of duplicates, keep the one with highest MAF
# MAF is min(ALT_FREQS, 1-ALT_FREQS)
df['MAF'] = df['ALT_FREQS'].apply(lambda x: min(float(x), 1-float(x)))
to_keep = df.sort_values('MAF', ascending=False).drop_duplicates('base_id', keep='first')

# Write the list of variants to keep
to_keep['ID'].to_csv("${params.output_prefix}.keep.variants", index=False, header=False)

Create the final filtered dataset

plink2 --pfile ${input_prefix} \
       --extract ${variants_to_keep} \
       --make-pgen \
       --out ${params.output_prefix}

I generated this answer using amplicon.ai, a tool I've been building to make writing biofinformatics code easier. Feel free to try it out