Robust and scalable barcoding for massively parallel long-read sequencing

Abstract

Nucleic-acid barcoding is an enabling technique for many applications, but its use remains limited in emerging long-read sequencing technologies with intrinsically low raw accuracy. Here, we apply so-called NS-watermark barcodes, whose error correction capability was previously validated in silico, in a proof of concept where we synthesize 3840 NS-watermark barcodes and use them to asymmetrically tag and simultaneously sequence amplicons from two evolutionarily distant species (namely Bordetella pertussis and Drosophila mojavensis) on the ONT MinION platform. To our knowledge, this is the largest number of distinct, non-random tags ever sequenced in parallel and the first report of microarray-based synthesis as a source for large oligonucleotide pools for barcoding. We recovered the identity of more than 86% of the barcodes, with a crosstalk rate of 0.17% (i.e., one misassignment every 584 reads). This falls in the range of the index hopping rate of established, high-accuracy Illumina sequencing, despite the increased number of tags and the relatively low accuracy of both microarray-based synthesis and long-read sequencing. The robustness of NS-watermark barcodes, together with their scalable design and compatibility with low-cost massive synthesis, makes them promising for present and future sequencing applications requiring massive labeling, such as long-read single-cell RNA-Seq.

HycDemux: a hybrid unsupervised approach for accurate barcoded sample demultiplexing in nanopore sequencing

“…In these datasets, D1 D3 integrate the official nanopore barcode. In order to evaluate the robustness of the demultiplexing method and the performance of demultiplexing a large number of non-ONT barcodes [ 44 , 45 ], we increased the number of barcodes and generated simulated multi-sample sequencing data (D4 D7 and DB4 DB7), these barcodes were randomly generated, and a certain edit distance was guaranteed (edit distance = 14.5 ± 1.8). The barcode consist of three primary components: upstream flanking region, variable region, and downstream flanking region (Fig.…”

Melon: metagenomic long-read-based taxonomic identification and quantification using marker genes

“…We did not use the theoretical relative abundances provided by ZymoBIOMICS as they might differ from the expected ones due to mishandling during library preparation (Supplementary Figure 1) [38]. Reads that mapped to the reference genomes of yeasts or remained unmapped were discarded to avoid overestimation in genome copies caused by eukaryotes [39], and potential contamination by e.g., barcode "crosstalk" [40].…”

Fast trimer statistics facilitate accurate decoding of large random DNA barcode sets even at large sequencing error rates

“…In the most favorable case of independent random errors, three or more errors can occur frequently; burst errors such as stuttering or repeated deletions only make things worse. This paper explores a possible solution via the use of random barcodes ("randomers") [30], that is, barcode libraries of any desired size (≳ 10 6 , for example) whose codewords are approximately uniformly random, as generated by computer, with constraints of GC content, homopolymers, etc., easily imposed. Like "designed" barcodes, random barcodes would be synthesized in defined oligo pools, but with the difference that the number of pools could be as large as desired.…”