Amplification-free Illumina sequencing-library preparation facilitates improved mapping and assembly of (G+C)-biased genomes

Abstract

Amplification artifacts introduced during library preparation for the Illumina Genome Analyzer increase the likelihood that an appreciable proportion of these sequences will be duplicates and cause an uneven distribution of read coverage across the targeted sequencing regions. As a consequence, these unfavorable features result in difficulties in genome assembly and variation analysis from the short reads, particularly when the sequences are from genomes with base compositions at the extremes of high or low G+C content. Here we present an amplification-free method of library preparation, in which the cluster amplification step, rather than the PCR, enriches for fully ligated template strands, reducing the incidence of duplicate sequences, improving read mapping and single nucleotide polymorphism calling and aiding de novo assembly. We illustrate this by generating and analyzing DNA sequences from extremely (G+C)-poor (Plasmodium falciparum), (G+C)-neutral (Escherichia coli) and (G+C)-rich (Bordetella pertussis) genomes.

Figure 1. No-PCR library preparation

In both standard and no-PCR library preps, partially complementary (‘Y-shaped’) adapters with a 3′ T overhang are ligated onto fragmented, end-repaired, 3′ A-tailed DNA. Whereas standard adapters consist only of sections to which read 1 and read 2 sequencing primers hybridize (R1 and R2′), no-PCR adapters also contain sequences that facilitate hybridization to oligonucleotides attached to the flowcell surface (FP1 and FP2′). The standard library prep uses PCR to add these sections, and to enrich for fully ligated templates which then amplify on the flow cell surface. With the no-PCR approach, the flowcell itself is used to select for fully ligated template molecules. All no-PCR templates hybridise to the flowcell in the same orientation, because only the FP2′ sequence is reverse complementary to a flowcell oligonucleotide.

Figure 2. Distribution of genome sequence coverage

The distribution of sequence coverage across the unmasked genomes are shown for various datasets with (STD) or without (NP) the PCR step. (a) % of unmasked genome against depth of genome base coverage and (b) accumulated % of unmasked genome against depth of genome base coverage for malaria strains (P. falciparum (PF) 2,3,88 and 85 and 3D7) with either long (L) or short (S) reads. (c) % of unmasked genome against depth of genome base coverage and (d) accumulated % of unmasked genome against depth of genome base coverage for E. coli 042 and B. pertussis ST24

Figure 3. Distribution of sequenced reads for different values of GC content

GC profiles for raw and mapped sequence data for the malaria strains NP-3D7-S and STD-PF2 are shown alongside simulated data (‘Shred-3D7′) for comparison. GC levels are calculated in a window size of read length and therefore the peak of fraction reads is dependent upon read length. A shift away from the simulated data curve, towards a more balanced GC composition is evident for the STD-PF2 sequence data, indicating severe bias.

Figure 4. Frequencies of duplicate sequences

Percentage of matched reads against duplication depth for sequence data derived from libraries prepared both with and without a PCR step. (a) Duplication frequencies for Plasmodium libraries; (b) duplication frequencies for E. coli and B. pertussis libraries.