Insight into biases and sequencing errors for amplicon sequencing with the Illumina MiSeq platform.
Schirmer M, Ijaz UZ, D'Amore R, Hall N, Sloan WT, Quince C. · 2015
Read Insight into biases and sequencing errors for amplicon sequencing with the Illumina MiSeq platform. by Schirmer M, Ijaz UZ, D'Amore R, Hall N, Sloan WT, Quince C. (2015) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/europmc-pmc4381044.txt.
Introduction
Melanie Schirmer, Umer Ijaz, Rosalinda D'Amore, Neil Hall, William Sloan, and Christopher Quince mapped the errors of a popular DNA sequencing machine, in Nucleic Acids Research in 2015. The Illumina MiSeq reads DNA fast and cheaply, and small labs now use it for gene surveys and microbe studies. Yet its error patterns were poorly known, which risks false claims about real genetic variation. The team ran a large test using gene-marker sequencing data from microbial samples.
How the sample library was prepared, and which primers were chosen, proved the biggest sources of bias, each leaving distinct error patterns. The team also tested cleanup strategies head to head. Trimming low-quality ends, correcting errors, then overlapping paired reads cut substitution errors by 93 percent on average. Their recipe gives labs a firmer base for telling real variation from machine noise.
Catalog record
| CREATOR | Schirmer M, Ijaz UZ, D'Amore R, Hall N, Sloan WT, Quince C. |
|---|---|
| YEAR | 2015 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 53 KB |
| LICENSE | CC BY 4.0 |
| RIGHTS | CC BY |
| SOURCE | Europe PMC |
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Insight into biases and sequencing errors for amplicon sequencing with the Illumina MiSeq platform By Melanie Schirmer, Umer Z Ijaz, Rosalinda D'Amore, Neil Hall, William T Sloan, Christopher Quince Published in: Nucleic Acids Research DOI: 10.1093/nar/gku1341 ABSTRACT Abstract With read lengths of currently up to 2 × 300 bp, high throughput and low sequencing costs Illumina's MiSeq is becoming one of the most utilized sequencing platforms worldwide. The platform is manageable and affordable even for smaller labs. This enables quick turnaround on a broad range of applications such as targeted gene sequencing, metagenomics, small genome sequencing and clinical molecular diagnostics. However, Illumina error profiles are still poorly understood and programs are therefore not designed for the idiosyncrasies of Illumina data. A better knowledge of the error patterns is essential for sequence analysis and vital if we are to draw valid conclusions. Studying true genetic variation in a population sample is fundamental for understanding diseases, evolution and origin. We conducted a large study on the error patterns for the MiSeq based on 16S rRNA amplicon sequencing data. We tested state-of-the-art library preparation methods for amplicon sequencing and showed that the library preparation method and the choice of primers are the most significant sources of bias and cause distinct error patterns. Furthermore we tested the efficiency of various error correction strategies and identified quality trimming (Sickle) combined with error correction (BayesHammer) followed by read overlapping (PANDAseq) as the most successful approach, reducing substitution error rates on average by 93%. INTRODUCTION The announcement by Roche to withdraw the GS FLX 454 pyrosequencing platform emphasizes the need for a better understanding of Illumina errors. 454 and Illumina sequencing errors are fundamentally different and require different strategies with regard to the downstream analysis. The majority of errors in 454 data are related to homopolymers (1,2). For Illumina, on the other hand, substitution type miscalls are the dominant source of errors. The Illumina sequencing technology is based on array formation. The sequencing templates are immobilized on a flow cell and a subsequent solid-phase bridge amplification generates up to 1000 copies in close proximity (cluster generation). The sequencing-by-synthesis technology uses fluorescently labelled reversible terminator-bound dNTPs (deoxyribonucleotides: A,C,G,T) for the polymerization. Only one base is added in each step due to the 3′ termination of the incorporated nucleotide. The fluorophores are illuminated by a red laser for A and C and a green laser for G and T and imaged through different filters to identify the four different nucleotides. The fluorescent labels and the 3′ terminators are then removed in order for the next cycle to commence. Challenges arise due to a strong correlation of A and C as well as G and T intensities as a result of similar emission spectra of the fluorophores and limitations of the filters that are used to separate the signals. Furthermore problems known as phasing and pre-phasing can cause noise in the cluster signal. Phasing can occur due to problems with the enzyme kinetics such as incomplete removal of the 3′ terminators or the fluorophores and cause the synthesis of some molecules in a cluster to lag behind. During pre-phasing on the other hand the synthesis advances too fast which can be caused by inadequate flushing of the flow cell, by sequences in a cluster skipping an incorporation cycle or the incorporation of nucleotides without an effective 3′ terminator. The number of affected sequences increases with each cycle and thus limits the overall read length. Overall, substitution type miscalls are the major source of errors for Illumina sequencing (3). Most previous studies on Illumina-specific errors have concentrated on the Genome Analyzer (GAII) and the HiSeq 2000 (4–6). Significant improvements in the technology and software have generally improved error rates but we still face systematic errors in Illumina sequencing. Nakamura et al. (7) identified two sequence patterns in Illumina GAII data that trigger errors during the sequencing process—firstly, inverted repeats and secondly, GGC sequences. They suspect that the first pattern causes dephasing by inhibiting single-base elongation through folding of single-stranded DNA and that the second pattern causes altered enzyme preference on the lagging-strand. They also report that mismatches are mainly observed in reads sequenced in the same direction and there is a strong correlation between average base call quality and mismatch rate. Their study also showed that the GGC pattern does not always trigger a sequencing error. The pattern ‘may occur once every 64 bases by chance’ (7) but sequence-specific errors (SSEs) are less common. Furthermore a significant number of SSE positions were not associated with the identified sequence patterns suggesting that other factors may be a significant cause for sequencing errors. For our experiments we used a variety of single species samples as well as a complex mock community consisting of 59 organisms. We developed a program that enables us to infer error profiles based on sequencing data from mock communities. Erroneous base calls are identified based on the alignment of the reads against the known reference genomes. Our software can identify mismatches, insertions and deletions for any sequenced mock data set. This allowed us to study and compare the impact of: library preparation methods, run, input DNA amount, number of polymerase chain reaction (PCR) cycles, Taq, DNA template and forward and reverse primer combinations. We provide an in-depth analysis of the errors occurring on both read directions for all types of errors and tested the reliability of quality scores. It has been reported previously that the per-base quality scores can be inaccurate and co-variation has been observed with attributes like sequencing technology, machine cycle and sequence context (8). We will show that the accuracy of the quality scores varies depending on which library preparation method was used. The differentiation of true variation and context-specific sequencing errors is a major challenge in next generation sequencing (NGS) analysis. Being able to infer error profiles for individual sequencing runs has the potential to greatly improve our ability to correct errors and thus enhance further sequencing analysis. Currently available programs aiming to address errors in Illumina sequencing data include quality trimming, error correction and read overlapping. During quality trimming the average quality scores are computed over a sliding window across the whole read. The start of the read is trimmed until the average quality score is larger than a certain threshold and the end of the read is trimmed if the average quality score falls below the threshold. Error correction will be tested with the program BayesHammer. During this approach a Hamming graph is constructed based on the k-mer composition of the reads. Using Bayesian subclustering the connected components are further divided into subclusters taking the quality values into account. For each subcluster the central k-mer is assumed to be the true sequence and the remaining sequences are corrected accordingly. Errors can also be corrected by overlapping paired-end reads. The reads are aligned and the optimal overlap is determined followed by error correction and assembling of the reads into a single sequence. Here, the quality scores are used for aligning the reads as well as for the error correction. We will test the capacity of these approaches and discuss their limitations. MATERIALS AND METHODS Mock community and sequencing data We sequenced a variety of