Vcftools

# 1) SNP-only filtered VCF
vcftools \
  --gzvcf cohort.vcf.gz \
  --remove-indels \
  --recode --recode-INFO-all \
  --out cohort.snps_only

# 2) Common-variant, low-missingness filter
vcftools \
  --gzvcf cohort.vcf.gz \
  --maf 0.05 \
  --max-missing 0.9 \
  --recode --recode-INFO-all \
  --out cohort.filtered

# 3) Allele-frequency report for one chromosome
vcftools \
  --gzvcf cohort.vcf.gz \
  --freq \
  --chr 1 \
  --out chr1.freq

# 4) Site mean depth summary
vcftools \
  --gzvcf cohort.vcf.gz \
  --site-mean-depth \
  --out cohort.depth

Recommended Workflow

1. Pick the correct input mode first: --vcf, --gzvcf, or --bcf.
2. Separate filtering passes from statistics passes so outputs stay interpretable.
3. When writing a new VCF, use --recode and usually --recode-INFO-all.
4. Keep the filtering thresholds documented, because small MAF or missingness changes can materially alter results.

Guardrails

• If --out is omitted, files default to the out.* prefix in the current directory.
• --recode writes a new VCF; without it many filtering commands only emit summary files.
• --max-missing 1.0 keeps only sites with no missing genotypes, which is often much harsher than intended.
• Distinguish site filters from genotype-aware downstream interpretation; filtering can strongly bias frequency-based summaries.