VCF and bfile sample IDs mismatch

FAQ: VCF and bfile sample IDs do not match

Problem

This may occur when sample IDs were anonymised during data upload to an imputation server. Then, the IDs in the VCF files do not match the IIDs in the PLINK .fam file, and PLINK may report that no samples remain after --keep in 01g.

Solution

Create a two-column mapping file without a header:

OLD_ID_1 NEW_ID_1
OLD_ID_2 NEW_ID_2

Then update the VCF IDs with PLINK 2:

for chr in {1..22}; do
    plink2 \
        --vcf "chr${chr}.vcf.gz" \
        --update-ids id_map.txt \
        --export vcf bgz id-paste=iid \
        --out "chr${chr}_ids_updated"
done

The id-paste=iid option ensures that only the IID is written to the VCF header, rather than a combined FID_IID value. This is because a VCF file can only store each ID in one column.

Check the updated IDs before rerunning the pipeline:

zgrep -m1 '^#CHROM' chr1_ids_updated.vcf.gz | less -S

The line should contain the IIDs your bfiles have:

#CHROM POS ID REF ALT QUAL FILTER INFO FORMAT SAMPLE_1 SAMPLE_2 ...

PLINK 2 documentation:

• [Updating sample IDs]