How to optimize the precision of allele and haplotype frequency estimates using pooled-sequencing data - Supplementary material S3: Effects of average sequencing depth and overdispersion on haplotype frequency estimates. Frequency estimation using pool sequencing

Eleven inbred lines from three distant Mediterranean populations (7 from Port La Nautique, France ; 2 from Centuri, Corsica and 2 from Alfarnate, Spain) were chosen from predominantly selfing populations with low heterozygosity (< 3% in all three populations). In 2014, all accessions were grown in the greenhouse, for two months under short-day conditions before sampling leaves. We weighed and pooled leaves from three ("Pool3", Table S2) or eight ("Pool8", Table S2) different accessions. For each of the 11 accessions and each of the two pools of leaves, DNA was extracted following the protocol detailed in Loridon et al. (2013). Each of the two DNA pools was further divided into four replicates for sequencing, resulting in a total of eight pooled samples. For each of the 11 individual or eight pooled samples, 200ng of genomic DNA was digested using one of two restriction enzyme (ApeKI or ECOT22I, New England Biolabs, Ipswich, MA, USA) to prepare paired-end libraries (Elshire et al. 2011). After ligation of barcoded adapters, DNA libraries were purified with AMPure (Agencourt Bioscience Corporation, Beckman Coulter, Beverly, Massachusetts, USA), pooled in approximately equal amounts, jointly size-selected for a mean insert size of 490 bp (ApeKI-digested samples) or 440 bp (ECOT22I-digested samples) on pre-cast agarose gel cassettes (Blue Pippin, Sage Science Inc., Beverly, Massachusetts, USA) and amplified with 18 PCR cycles prior to sequencing. We sequenced all libraries with 250 bp paired-end reads with moderate sequencing depth (λ ~10 X) on an Illumina MiSeq.