WIXLIB

Molecular Barcodes - FAQDetect low allele frequency variants with HaloPlexHSAuthorsQ&A IntroductionAnniek De Witte1, Katie L. Zobeck1,Linus Forsmark , Christian LeCocq ,11Heather Tao1, Bahram Arezi2 , MagnusIsaksson11Agilent Technologies, Inc. Santa Clara,CA, USA. 2 Agilent Technologies, Inc.La Jolla, CA, USAWhat is NGS?Next generation sequencing (NGS), massively parallel sequencing, and high-throughputsequencing are related terms that describe an innovation in DNA sequencing technologythat allows for the sequencing of millions of small fragments of DNA at the same time.The result of this modern technology is a massive increase in the number of base pairssequenced compared to the standard Sanger sequencing method.Among the most popular applications using NGS are genomic DNA variant analysis andRNA expression analysis. The extent of these analyses can be as wide as the wholegenome (WGS or whole genome sequencing) and whole exome (WES or whole exomesequencing), or as focused as specific regions and gene panels.While WGS may capture all possible mutations, WES and in particular targeted sequencing,remains advantageous for achieving very high coverage of the regions of interest whilestaying cost-effective and keeping the complexity of data interpretation manageable.Having very high sequencing coverage is especially important for discovering cancermutations present at low fractions (1).How can a target region be enriched before NGS?There are several ways to enrich regions of interest before NGS. The two most commonlyused approaches are: 1) hybridization capture from sequencing libraries using targetspecific probes and 2) PCR amplification directly from sample DNA using target-specificprimers (2, 3).With hybrid capture methods, DNA fragments are hybridized in solution to sequencespecific capture probes corresponding to targeted regions of the genome. Examples ofhybridization capture technology include Agilent SureSelect, NimbleGen SeqCap, andIllumina TruSeq (4).Amplicon sequencing enriches target genes by PCR with a set of primers for the exonsof selected genes prior to NGS. Examples of amplicon capture technology include purePCR-based methods such as Ion Torrent AmpliSeq, RainDance ThunderBolts or IlluminaTruSeq amplicon, and hybridization and extension methods such as Agilent HaloPlexHS (5).

What are potential sources oferrors in NGS?Advancements in NGS and targetenrichment have made it possible todetect low-occurrence mutations in aheterogeneous mixture. However, furtheradvancements are hindered by systemicerrors in PCR and sequencing methods.Library preparation, target enrichment,and sequencing all use DNA polymeraseand amplification steps. These processesintroduce bias in the form of duplicatesand associated non-uniform amplificationand artifacts defined as polymerase errorsthat generate sequence changes notpresent in the original sample (1). To putit in numbers, with the commonly usedIllumina sequencing instruments, theerror rate can vary from 1% to 0.05%,depending on factors such as the readlength, base-calling algorithms used, andthe type of variants detected (6). Molecularbarcodes can enhance confidence,particularly in clinical research sampleswherein the mutation prevalence can be inthe same range as the error rate.Q&A Molecular BarcodesWhat are molecular barcodes?The concept of molecular barcodes isthat each original DNA fragment, withinthe same sample, is attached to a uniquesequence barcode. Molecular barcodesare usually designed as a string of totallyrandom nucleotides (such as NNNNNNN),partially degenerate nucleotides (such asNNNRNYN), or defined nucleotides (whentemplate molecules are limited). Molecularbarcodes allow for the accountingof sequencer and PCR errors in highcoverage NGS data.What is the difference betweenmolecular barcodes and samplebarcodes?It is common to have both samplebarcodes and molecular barcodes inthe same sequencing reads. Molecularbarcodes can be used to reduce falsepositives. Whereas sample barcodes, alsocalled indexed adaptors, are customarilyused in most current NGS workflowsand allow the mixing of samples priorto sequencing. These sample barcodesact like identifiers or tags that let youdetermine from which sample the readoriginated, enabling the assaying ofmultiple samples in a single run. Samplebarcodes are typically short specificsequences that are incorporated prior toDNA sequencing. The barcodes will besequenced together with the unknownsample DNA. After sequencing thereads are sorted by barcode and groupedtogether (demultiplexing).Why do molecular barcodes matter?The concept of molecular barcodesis that each fragment is attached to aunique molecular barcode. Sequencereads that have different molecularbarcodes represent different original DNAmolecules, while reads that have the samebarcodes are the result of PCR duplicationfrom the same original molecule.Molecular barcodes do not prevent PCRduplication from happening but theyallow users to track the duplicates andremove them from the downstreamanalysis. Additionally, by using molecularbarcodes PCR artifacts can be separatedfrom real variants present in the originalmolecules, allowing for variant detectionat a much lower variant allele frequency(VAF). Peng et al. (1) developed an NGStarget enrichment process that integratesmolecular barcodes into high multiplexPCR amplicon sequencing. They compareddata with and without molecular barcodeand showed that with high sensitivitysettings, using molecular barcodessignificantly reduced false positives.Do hybridization capture methodsbenefit from molecular barcodes?Although random shearing beforehybridization capture creates random anddiverse fragment ends that can be usedas unique identifiers for each startingmolecule, molecular barcodes are stilluseful in hybridization-based targetenrichment. First, at very high sequencingread depths (10,000 - 50,000X raw reads)it becomes clear that the fragmentscreated by random shearing are actuallynot completely random (7). Secondly dueto concerns about cost, throughput, andyield, the field of NGS is moving towardsreplacing random shearing with enzymaticfragmentation before hybridization2based target enrichment. This enzymaticshearing leads to much less randomfragment ends and also hampersthe ability to track different startingmolecules and to remove PCR duplicatesand associated amplification artifacts.What is the history of molecularbarcodes?Tagging individual templates with amolecular barcode has been proposedand reported since 2007 to alleviate theproblems of PCR duplication and biasedamplification in NGS. In the past, molecularbarcodes or molecular indexes have beengiven various names such as uniqueidentifiers (UID), unique molecular identifiers(UMI), primer ID, or duplex barcodes.Molecular barcodes can be addedto a template using several differenttechniques. One approach is to incorporatethem into the sequencing adaptors duringthe library construction step for genomesequencing. This “duplex sequencing”greatly reduces errors by independentlytagging and sequencing each of the twostrands of a DNA duplex (8). Anotherapproach is to incorporate the barcodesinto molecular inversion probes fortargeted somatic mutation detection. WithsmMIP (for single molecule molecularinversion probes) single-molecule taggingis combined with multiplex targetedcapture to enable practical and highlysensitive detection of low variant allelefrequency (9). Finally, molecular barcodescan also be incorporated into targetspecific PCR primers, usually in the form ofa short stretch of random bases.How are HaloPlexHS molecularbarcodes created?HaloPlexHS is a combination of ampliconbased and target enrichment sequencing.The protocol utilizes specificity gainedfrom restriction enzyme recognition,hybridization and DNA ligation to capturemolecules originating from the targetregion to be sequenced. To enableidentification of duplicate reads fromlibraries prepared with HaloPlexHS, wehave added a molecular barcode to theintroduced primer vector (Figure 1). TheHaloPlexHS workflow tags one barcode perfragment. In the case when HaloPlexHS

A1Digest and denature sample DNATarget RegionLigate and captureuniquely barcoded targets3Streptavidin2Hybridize probe library to DNA targets4Amplify enriched fragments by PCRPCR primersBiotinSequencing primer motifIndexBridge or emulsionPCR primerTarget & complementaryprobe sequenceMolecular barcodeB IlluminaI2Molecularbarcode, 10 ntR1Vectorsequence, 1 ntR2Ion PGMI1Samplebarcode, 8 ntMolecularbarcode, 10 ntIonXpressbarcodeVectorsequence, 15 ntFigure 1 A) Overview of the HaloPlexHS workflow. B) Schematic of HaloPlexHS libraries. For librariessequenced on the Illumina platform, the molecular barcodes are introduced as part of Illumina's dualindexing system. For libraries sequenced on the Ion PGM, the barcodes are incorporated at the beginningof the reads. Libraries are compatible with standard sequencing protocols.3captures both strands (i.e. when using theFFPE design option in SureDesign), bothdouble-stranded molecules are taggedindependently. The molecular barcodeconsists of ten degenerate bases allowingfor over one million unique sequences tobe present for tagging of molecules. Thebarcodes allow you to differentiate truevariants from PCR or sequencing errors.What level of coverage do I need totarget for the sensitivity I require?Combining high-fidelity DNA polymeraseswith the power of molecular barcodes todistinguish errors from true mutationsmakes it possible in theory to detectmutations at 1% or lower variant allelefrequency. In reality, however, undersampling and sampling bias challenges thepractical application of molecular barcodesand deep sequencing (10). To accuratelyassess the mutant representation in a poolof templates, all the templates within thepool have to be uniformly tagged, with noerror occurring during the tagging process,and the tags have to be immune tomutations in subsequent amplification.In an online article, Jon Armstrong (11)discussed the balance between theamount (or copies) of DNA going intothe library process, sequence coveragegenerated and the lower theoreticalthreshold of minor allele frequencydetection. If we want to detect variantsconfidently to 0.1% sensitivity whilerequiring 4 reads to show the alternateallele, we would require 4,000X coverageafter deduplication (4 reads / 0.001). Ina perfect world, we would need at least4,000 haploid genome copies, or 2000cells, or 12 ng of DNA. In other words,no matter how many sequencing readsyou generate, you can never attain agreater deduplicated coverage than thenumber of genomics copies going ontothe sequencer. However, these numbersincrease even further due to process noise,variability, PCR amplifications, and lossduring the molecular procedures.