Palindromic sequence artifacts generated during next generation sequencing library preparation from historic and ancient DNA

Degradation-specific processes and variation in laboratory protocols can bias the DNA sequence composition from samples of ancient or historic origin. Here, we identify a novel artifact in sequences from historic samples of Atlantic cod (Gadus morhua), which forms interrupted palindromes consisting of reverse complementary sequence at the 59 and 39-ends of sequencing reads. The palindromic sequences themselves have specific properties - the bases at the 59-end align well to the reference genome, whereas extensive misalignments exists among the bases at the terminal 39-end. The terminal 39 bases are artificial extensions likely caused by the occurrence of hairpin loops in single stranded DNA (ssDNA), which can be ligated and amplified in particular library creation protocols. We propose that such hairpin loops allow the inclusion of erroneous nucleotides, specifically at the 39-end of DNA strands, with the 59-end of the same strand providing the template. We also find these palindromes in previously published ancient DNA (aDNA) datasets, albeit at varying and substantially lower frequencies. This artifact can negatively affect the yield of endogenous DNA in these types of samples and introduces sequence bias.