Phylogenetic placement of short reads
To further compare the primer biases at the sequence level, we used the Rapid Axelerated Maximum Likelihood – Evolutionary Placement Algorithm (RAxML-EPA; Berger, Krompass, & Stamatakis, 2011) approach to ascertain the phylogenetic identities of the three abundant and important phytoplanktonic groups in the Arctic, namely Chlorophyta, Bacillariophyceae and Haptophyceae. To do this, phylogenetic reference trees were generated for each taxonomic group using Randomized Axelerated Maximum Likelihood (RAxML v.8; Stamatakis, 2014) based on literature (Lovejoy et al., 2007; Liu, Aris-Brosou, Probert, & de Vargas, 2010). HTS-generated OTUs were first searched against the NCBI GenBank using BLASTn to identify the best hits. Most of the highly similar sequences (> 98%, e-value 1e-10) were selected and aligned in MUSCLE (Edgar, 2010), with the alignment manually edited and trimmed in MEGA v.7.0 (Kumar, Stecher, & Tamura, 2016), and used to build robust clade-specific reference trees with RAxML v.8 using the GTRGAMMA model, run 1000 times (Stamatakis, 2014). Reference trees were generated separately for each taxonomic group. Short OTUs were then aligned with the reference sequences and mapped back onto the reference trees using the -n option in the EPA-RAxML v.8.2 (Berger et al., 2011). Trees generated from each primer pair were then visualized in Dendroscope v.3.0 (Huson & Scornavacca, 2012) and compared with each other.