Fig. 3

Examples of genotype pooling simulation at the block level. a Configuration with 1 sample carrying the minor allele. This carrier is identified after pooling, but not if it has a heterozygous (1) or a minor homozygous (2) genotype. b Configuration with 2 samples carrying the minor allele. At least 2 of the 4 samples highlighted in grey are minor allele carriers, but the genotypes of these 4 samples are indeterminate. The first step is encoding and pooling. Encoding assigns every sample to a pool and defines its pool coordinates. For instance in a, the sample at the top-left corner of the matrix has coordinates (1, 5). Pooling computes the genotype of a pool as if its would tested on a SNP-chip. Pool 5 (P5, most left) has genotype 1: both alleles 0 and 1 are detected among the samples. Pool 1 has genotype 0 because only the allele 0 is detected. The decoding step infers the pooled genotype of each sample from the genotypes of its coordinates. The genotype can be \(-1\) i.e. indeterminate when both coordinates have genotype 1, or fully determined else. In subfigure 3a, the sample with coordinates (3, 5) carries the alternate allele, but there can be 1 or 2 copies of it. \(\psi\) is the observed pooling pattern that results from grouped genotyping, given as the number of row- and column-pools having the genotypes (0, 1, 2). In the example 3a, there are 3 row-pools having genotype 0, 1 row-pool having genotype 1 and 0 having genotype 2, likewise for the column-pools. c and d Simulation example of genotype pooling and imputation outcomes for markers from the 1KGP data (chromosome 20). The genotypes are represented as unphased GT. From top to bottom: true genotype data, pooled genotypes, imputed genotypes. c SNP 20:264365, \(MAF = 0.4625\). d SNP 20:62915126, \(MAF = 0.00625\)