Identification and quantification of chimeric sequencing reads in a highly multiplexed RAD-seq protocol

Abstract

Highly multiplexed approaches have become a common practice in genomic studies. They have improved the cost-effectiveness of genotyping hundreds of individuals by using combinatorially-barcoded adapters. These strategies, however, can potentially misassign reads to incorrect samples. Here we used a modified quaddRAD protocol to analyse the occurrence of index hopping and PCR chimeras in a series of experiments with up to a 100 multiplexed samples per sequencing lane (total n = 639). We created two types of sequencing libraries: four libraries of Type A, where PCR reactions were run on individual samples before multiplexing, and three libraries of Type B, where PCRs were run on pooled samples. We used fixed pairs of inner barcodes to identify chimeric reads. Type B libraries show a higher percentage of misassigned reads (1.15%) compared to Type A libraries (0.65%). We also quantify the commonly undetectable chimeric sequences that occur whenever multiplexed groups of samples with different outer barcodes are sequenced together on a single flow cell. Our results suggest that these types of chimeric sequences represent up to 1.56% and 1.29% of reads in Type A and B libraries, respectively. We review the source of such errors, provide recommendations for developing highly-multiplexed RAD-seq protocols and analysing the resulting data to minimise the generation of chimeric sequences, allow their quantification, and provide finer control over the number of PCR cycles necessary to generate enough input DNA for library preparation.