Evaluation of the Ion Torrent Personal Genome Machine for Gene-Targeted Studies Using Amplicons of the Nitrogenase Gene nifH

The sequencing chips and kits of the Ion Torrent Personal Genome Machine (PGM), which employs semiconductor technology to measure pH changes in polymerization events, have recently been upgraded. The quality of PGM sequences has not been reassessed, and results have not been compared in the context of a gene-targeted microbial ecology study. To address this, we compared sequence profiles across available PGM chips and chemistries and with 454 pyrosequencing data by determining error types and rates and diazotrophic community structures. The PGM was then used to assess differences in nifH-harboring bacterial community structure among four corn-based cropping systems. Using our suggested filters from mock community analyses, the overall error rates were 0.62, 0.36, and 0.39% per base for chips 318 and 314 with the 400-bp kit and chip 318 with the Hi-Q chemistry, respectively. Compared with the 400-bp kit, the Hi-Q kit reduced indel rates by 28 to 59% and produced one to seven times more reads acceptable for downstream analyses. The PGM produced higher frameshift rates than pyrosequencing that were corrected by the RDP FrameBot tool. Significant differences among platforms were identified, although the diversity indices and overall site-based conclusions remained similar. For the cropping system analyses, a total of 6,182 unique NifH operational taxonomic units at 5% amino acid dissimilarity were obtained. The current crop type, as well as the crop rotation history, significantly influenced the composition of the soil diazotrophic community detected.     

Aberration-corrected ultrafine analysis of miRNA reads at single-base resolution: a k-mer lattice approach

Raw sequencing reads of miRNAs contain machine-made substitution errors, or even insertions and deletions (indels). Although the error rate can be low at 0.1%, precise rectification of these errors is critically important because isoform variation analysis at single-base resolution such as novel isomiR discovery, editing events understanding, differential expression analysis, or tissue-specific isoform identification is very sensitive to base positions and copy counts of the reads. Existing error correction methods do not work for miRNA sequencing data attributed to miRNAs’ length and per-read-coverage properties distinct from DNA or mRNA sequencing reads. We present a novel lattice structure combining kmers, (k – 1)mers and (k + 1)mers to address this problem. The method is particularly effective for the correction of indel errors. Extensive tests on datasets having known ground truth of errors demonstrate that the method is able to remove almost all of the errors, without introducing any new error, to improve the data quality from every-50-reads containing one error to every-1300-reads containing one error. Studies on experimental miRNA sequencing datasets show that the errors are often rectified at the 5′ ends and the seed regions of the reads, and that there are remarkable changes after the correction in miRNA isoform abundance, volume of singleton reads, overall entropy, isomiR families, tissue-specific miRNAs, and rare-miRNA quantities.     

Performance comparison of benchtop high-throughput sequencing platforms

Three benchtop high-throughput sequencing instruments are now available. The 454 GS Junior (Roche), MiSeq (Illumina) and Ion Torrent PGM (Life Technologies) are laser-printer sized and offer modest set-up and running costs. Each instrument can generate data required for a draft bacterial genome sequence in days, making them attractive for identifying and characterizing pathogens in the clinical setting. We compared the performance of these instruments by sequencing an isolate of Escherichia coli O104:H4, which caused an outbreak of food poisoning in Germany in 2011. The MiSeq had the highest throughput per run (1.6 Gb/run, 60 Mb/h) and lowest error rates. The 454 GS Junior generated the longest reads (up to 600 bases) and most contiguous assemblies but had the lowest throughput (70 Mb/run, 9 Mb/h). Run in 100-bp mode, the Ion Torrent PGM had the highest throughput (80–100 Mb/h). Unlike the MiSeq, the Ion Torrent PGM and 454 GS Junior both produced homopolymer-associated indel errors (1.5 and 0.38 errors per 100 bases, respectively).     

Pollux: platform independent error correction of single and mixed genomes

Background

Second-generation sequencers generate millions of relatively short, but error-prone, reads. These errors make sequence assembly and other downstream projects more challenging. Correcting these errors improves the quality of assemblies and projects which benefit from error-free reads.

Results

We have developed a general-purpose error corrector that corrects errors introduced by Illumina, Ion Torrent, and Roche 454 sequencing technologies and can be applied to single- or mixed-genome data. In addition to correcting substitution errors, we locate and correct insertion, deletion, and homopolymer errors while remaining sensitive to low coverage areas of sequencing projects. Using published data sets, we correct 94% of Illumina MiSeq errors, 88% of Ion Torrent PGM errors, 85% of Roche 454 GS Junior errors. Introduced errors are 20 to 70 times more rare than successfully corrected errors. Furthermore, we show that the quality of assemblies improves when reads are corrected by our software.

Conclusions

Pollux is highly effective at correcting errors across platforms, and is consistently able to perform as well or better than currently available error correction software. Pollux provides general-purpose error correction and may be used in applications with or without assembly.     

Correction of sequence-dependent ambiguous bases (Ns) from the 454 pyrosequencing system

Pyrosequencing of the 16S ribosomal RNA gene (16S) has become one of the most popular methods to assess microbial diversity. Pyrosequencing reads containing ambiguous bases (Ns) are generally discarded based on the assumptions of their non-sequence-dependent formation and high error rates. However, taxonomic composition differed by removal of reads with Ns. We determined whether Ns from pyrosequencing occur in a sequence-dependent manner. Our reads and the corresponding flow value data revealed occurrence of sequence-specific N errors with a common sequential pattern (a homopolymer + a few nucleotides with bases other than the homopolymer + N) and revealed that the nucleotide base of the homopolymer is the true base for the following N. Using an algorithm reflecting this sequence-dependent pattern, we corrected the Ns in the 16S (86.54%), bphD (81.37%) and nifH (81.55%) amplicon reads from a mock community with high precisions of 95.4, 96.9 and 100%, respectively. The new N correction method was applicable for determining most of Ns in amplicon reads from a soil sample, resulting in reducing taxonomic biases associated with N errors and in shotgun sequencing reads from public metagenome data. The method improves the accuracy and precision of microbial community analysis and genome sequencing using 454 pyrosequencing.     

Evaluation and optimisation of indel detection workflows for ion torrent sequencing of the BRCA1 and BRCA2 genes

Background

The Ion Torrent PGM is a popular benchtop sequencer that shows promise in replacing conventional Sanger sequencing as the gold standard for mutation detection. Despite the PGM’s reported high accuracy in calling single nucleotide variations, it tends to generate many false positive calls in detecting insertions and deletions (indels), which may hinder its utility for clinical genetic testing.

Results

Recently, the proprietary analytical workflow for the Ion Torrent sequencer, Torrent Suite (TS), underwent a series of upgrades. We evaluated three major upgrades of TS by calling indels in the BRCA1 and BRCA2 genes. Our analysis revealed that false negative indels could be generated by TS under both default calling parameters and parameters adjusted for maximum sensitivity. However, indel calling with the same data using the open source variant callers, GATK and SAMtools showed that false negatives could be minimised with the use of appropriate bioinformatics analysis. Furthermore, we identified two variant calling measures, Quality-by-Depth (QD) and VARiation of the Width of gaps and inserts (VARW), which substantially reduced false positive indels, including non-homopolymer associated errors without compromising sensitivity. In our best case scenario that involved the TMAP aligner and SAMtools, we achieved 100% sensitivity, 99.99% specificity and 29% False Discovery Rate (FDR) in indel calling from all 23 samples, which is a good performance for mutation screening using PGM.

Conclusions

New versions of TS, BWA and GATK have shown improvements in indel calling sensitivity and specificity over their older counterpart. However, the variant caller of TS exhibits a lower sensitivity than GATK and SAMtools. Our findings demonstrate that although indel calling from PGM sequences may appear to be noisy at first glance, proper computational indel calling analysis is able to maximize both the sensitivity and specificity at the single base level, paving the way for the usage of this technology for future clinical genetic testing.

Electronic supplementary material

The online version of this article (doi:10.1186/1471-2164-15-516) contains supplementary material, which is available to authorized users.     

Improving Indel Detection Specificity of the Ion Torrent PGM Benchtop Sequencer

The emergence of benchtop sequencers has made clinical genetic testing using next-generation sequencing more feasible. Ion Torrent''s PGM^TM is one such benchtop sequencer that shows clinical promise in detecting single nucleotide variations (SNVs) and microindel variations (indels). However, the large number of false positive indels caused by the high frequency of homopolymer sequencing errors has impeded PGM^TM''s usage for clinical genetic testing. An extensive analysis of PGM^TM data from the sequencing reads of the well-characterized genome of the Escherichia coli DH10B strain and sequences of the BRCA1 and BRCA2 genes from six germline samples was done. Three commonly used variant detection tools, SAMtools, Dindel, and GATK''s Unified Genotyper, all had substantial false positive rates for indels. By incorporating filters on two major measures we could dramatically improve false positive rates without sacrificing sensitivity. The two measures were: B-Allele Frequency (BAF) and VARiation of the Width of gaps and inserts (VARW) per indel position. A BAF threshold applied to indels detected by UnifiedGenotyper removed ∼99% of the indel errors detected in both the DH10B and BRCA sequences. The optimum BAF threshold for BRCA sequences was determined by requiring 100% detection sensitivity and minimum false discovery rate, using variants detected from Sanger sequencing as reference. This resulted in 15 indel errors remaining, of which 7 indel errors were removed by selecting a VARW threshold of zero. VARW specific errors increased in frequency with higher read depth in the BRCA datasets, suggesting that homopolymer-associated indel errors cannot be reduced by increasing the depth of coverage. Thus, using a VARW threshold is likely to be important in reducing indel errors from data with higher coverage. In conclusion, BAF and VARW thresholds provide simple and effective filtering criteria that can improve the specificity of indel detection in PGM^TM data without compromising sensitivity.     

Vindel: a simple pipeline for checking indel redundancy

Background

With the advance of next generation sequencing (NGS) technologies, a large number of insertion and deletion (indel) variants have been identified in human populations. Despite much research into variant calling, it has been found that a non-negligible proportion of the identified indel variants might be false positives due to sequencing errors, artifacts caused by ambiguous alignments, and annotation errors.

Results

In this paper, we examine indel redundancy in dbSNP, one of the central databases for indel variants, and develop a standalone computational pipeline, dubbed Vindel, to detect redundant indels. The pipeline first applies indel position information to form candidate redundant groups, then performs indel mutations to the reference genome to generate corresponding indel variant substrings. Finally the indel variant substrings in the same candidate redundant groups are compared in a pairwise fashion to identify redundant indels. We applied our pipeline to check for redundancy in the human indels in dbSNP. Our pipeline identified approximately 8% redundancy in insertion type indels, 12% in deletion type indels, and overall 10% for insertions and deletions combined. These numbers are largely consistent across all human autosomes. We also investigated indel size distribution and adjacent indel distance distribution for a better understanding of the mechanisms generating indel variants.

Conclusions

Vindel, a simple yet effective computational pipeline, can be used to check whether a set of indels are redundant with respect to those already in the database of interest such as NCBI’s dbSNP. Of the approximately 5.9 million indels we examined, nearly 0.6 million are redundant, revealing a serious limitation in the current indel annotation. Statistics results prove the consistency of the pipeline on indel redundancy detection for all 22 chromosomes. Apart from the standalone Vindel pipeline, the indel redundancy check algorithm is also implemented in the web server http://bioinformatics.cs.vt.edu/zhanglab/indelRedundant.php.

Electronic supplementary material

The online version of this article (doi:10.1186/s12859-014-0359-1) contains supplementary material, which is available to authorized users.     

Error and error mitigation in low-coverage genome assemblies

The recent release of twenty-two new genome sequences has dramatically increased the data available for mammalian comparative genomics, but twenty of these new sequences are currently limited to ～2× coverage. Here we examine the extent of sequencing error in these 2× assemblies, and its potential impact in downstream analyses. By comparing 2× assemblies with high-quality sequences from the ENCODE regions, we estimate the rate of sequencing error to be 1-4 errors per kilobase. While this error rate is fairly modest, sequencing error can still have surprising effects. For example, an apparent lineage-specific insertion in a coding region is more likely to reflect sequencing error than a true biological event, and the length distribution of coding indels is strongly distorted by error. We find that most errors are contributed by a small fraction of bases with low quality scores, in particular, by the ends of reads in regions of single-read coverage in the assembly. We explore several approaches for automatic sequencing error mitigation (SEM), making use of the localized nature of sequencing error, the fact that it is well predicted by quality scores, and information about errors that comes from comparisons across species. Our automatic methods for error mitigation cannot replace the need for additional sequencing, but they do allow substantial fractions of errors to be masked or eliminated at the cost of modest amounts of over-correction, and they can reduce the impact of error in downstream phylogenomic analyses. Our error-mitigated alignments are available for download.     

Decoding of superimposed traces produced by direct sequencing of heterozygous indels

Direct Sanger sequencing of a diploid template containing a heterozygous insertion or deletion results in a difficult-to-interpret mixed trace formed by two allelic traces superimposed onto each other. Existing computational methods for deconvolution of such traces require knowledge of a reference sequence or the availability of both direct and reverse mixed sequences of the same template. We describe a simple yet accurate method, which uses dynamic programming optimization to predict superimposed allelic sequences solely from a string of letters representing peaks within an individual mixed trace. We used the method to decode 104 human traces (mean length 294 bp) containing heterozygous indels 5 to 30 bp with a mean of 99.1% bases per allelic sequence reconstructed correctly and unambiguously. Simulations with artificial sequences have demonstrated that the method yields accurate reconstructions when (1) the allelic sequences forming the mixed trace are sufficiently similar, (2) the analyzed fragment is significantly longer than the indel, and (3) multiple indels, if present, are well-spaced. Because these conditions occur in most encountered DNA sequences, the method is widely applicable. It is available as a free Web application Indelligent at http://ctap.inhs.uiuc.edu/dmitriev/indel.asp.     

PCR-Induced Transitions Are the Major Source of Error in Cleaned Ultra-Deep Pyrosequencing Data

Background

Ultra-deep pyrosequencing (UDPS) is used to identify rare sequence variants. The sequence depth is influenced by several factors including the error frequency of PCR and UDPS. This study investigated the characteristics and source of errors in raw and cleaned UDPS data.

Results

UDPS of a 167-nucleotide fragment of the HIV-1 SG3Δenv plasmid was performed on the Roche/454 platform. The plasmid was diluted to one copy, PCR amplified and subjected to bidirectional UDPS on three occasions. The dataset consisted of 47,693 UDPS reads. Raw UDPS data had an average error frequency of 0.30% per nucleotide site. Most errors were insertions and deletions in homopolymeric regions. We used a cleaning strategy that removed almost all indel errors, but had little effect on substitution errors, which reduced the error frequency to 0.056% per nucleotide. In cleaned data the error frequency was similar in homopolymeric and non-homopolymeric regions, but varied considerably across sites. These site-specific error frequencies were moderately, but still significantly, correlated between runs (r = 0.15–0.65) and between forward and reverse sequencing directions within runs (r = 0.33–0.65). Furthermore, transition errors were 48-times more common than transversion errors (0.052% vs. 0.001%; p<0.0001). Collectively the results indicate that a considerable proportion of the sequencing errors that remained after data cleaning were generated during the PCR that preceded UDPS.

Conclusions

A majority of the sequencing errors that remained after data cleaning were introduced by PCR prior to sequencing, which means that they will be independent of platform used for next-generation sequencing. The transition vs. transversion error bias in cleaned UDPS data will influence the detection limits of rare mutations and sequence variants.     

Challenges with Using Primer IDs to Improve Accuracy of Next Generation Sequencing

Next generation sequencing technologies, like ultra-deep pyrosequencing (UDPS), allows detailed investigation of complex populations, like RNA viruses, but its utility is limited by errors introduced during sample preparation and sequencing. By tagging each individual cDNA molecule with barcodes, referred to as Primer IDs, before PCR and sequencing these errors could theoretically be removed. Here we evaluated the Primer ID methodology on 257,846 UDPS reads generated from a HIV-1 SG3Δenv plasmid clone and plasma samples from three HIV-infected patients. The Primer ID consisted of 11 randomized nucleotides, 4,194,304 combinations, in the primer for cDNA synthesis that introduced a unique sequence tag into each cDNA molecule. Consensus template sequences were constructed for reads with Primer IDs that were observed three or more times. Despite high numbers of input template molecules, the number of consensus template sequences was low. With 10,000 input molecules for the clone as few as 97 consensus template sequences were obtained due to highly skewed frequency of resampling. Furthermore, the number of sequenced templates was overestimated due to PCR errors in the Primer IDs. Finally, some consensus template sequences were erroneous due to hotspots for UDPS errors. The Primer ID methodology has the potential to provide highly accurate deep sequencing. However, it is important to be aware that there are remaining challenges with the methodology. In particular it is important to find ways to obtain a more even frequency of resampling of template molecules as well as to identify and remove artefactual consensus template sequences that have been generated by PCR errors in the Primer IDs.     

WaveNano: a signal-level nanopore base-caller via simultaneous prediction of nucleotide labels and move labels through bi-directional WaveNets

Background: The Oxford MinION nanopore sequencer is the recently appealing third-generation genome sequencing device that is portable and no larger than a cellphone. Despite the benefits of MinION to sequence ultra-long reads in real-time, the high error rate of the existing base-calling methods, especially indels (insertions and deletions), prevents its use in a variety of applications. Methods: In this paper, we show that such indel errors are largely due to the segmentation process on the input electrical current signal from MinION. All existing methods conduct segmentation and nucleotide label prediction in a sequential manner, in which the errors accumulated in the first step will irreversibly influence the final base-calling. We further show that the indel issue can be significantly reduced via accurate labeling of nucleotide and move labels directly from the raw signal, which can then be efficiently learned by a bi-directional WaveNet model simultaneously through feature sharing. Our bi-directional WaveNet model with residual blocks and skip connections is able to capture the extremely long dependency in the raw signal. Taking the predicted move as the segmentation guidance, we employ the Viterbi decoding to obtain the final base-calling results from the smoothed nucleotide probability matrix. Results: Our proposed base-caller, WaveNano, achieves good performance on real MinION sequencing data from Lambda phage. Conclusions: The signal-level nanopore base-caller WaveNano can obtain higher base-calling accuracy, and generate fewer insertions/deletions in the base-called sequences.     

Genetic evidence that both dNTP-stabilized and strand slippage mechanisms may dictate DNA polymerase errors within mononucleotide microsatellites

Mononucleotide microsatellites are tandem repeats of a single base pair, abundant within coding exons and frequent sites of mutation in the human genome. Because the repeated unit is one base pair, multiple mechanisms of insertion/deletion (indel) mutagenesis are possible, including strand-slippage, dNTP-stabilized, and misincorportion-misalignment. Here, we examine the effects of polymerase identity (mammalian Pols α, β, κ, and η), template sequence, dNTP pool size, and reaction temperature on indel errors during in vitro synthesis of mononucleotide microsatellites. We utilized the ratio of insertion to deletion errors as a genetic indicator of mechanism. Strikingly, we observed a statistically significant bias toward deletion errors within mononucleotide repeats for the majority of the 28 DNA template and polymerase combinations examined, with notable exceptions based on sequence and polymerase identity. Using mutator forms of Pol β did not substantially alter the error specificity, suggesting that mispairing-misalignment mechanism is not a primary mechanism. Based on our results for mammalian DNA polymerases representing three structurally distinct families, we suggest that dNTP-stabilized mutagenesis may be an alternative mechanism for mononucleotide microsatellite indel mutation. The change from a predominantly dNTP-stabilized mechanism to a strand-slippage mechanism with increasing microsatellite length may account for the differential rates of tandem repeat mutation that are observed genome-wide.     

Tool evaluation for the detection of variably sized indels from next generation whole genome and targeted sequencing data

Insertions and deletions (indels) in human genomes are associated with a wide range of phenotypes, including various clinical disorders. High-throughput, next generation sequencing (NGS) technologies enable the detection of short genetic variants, such as single nucleotide variants (SNVs) and indels. However, the variant calling accuracy for indels remains considerably lower than for SNVs. Here we present a comparative study of the performance of variant calling tools for indel calling, evaluated with a wide repertoire of NGS datasets. While there is no single optimal tool to suit all circumstances, our results demonstrate that the choice of variant calling tool greatly impacts the precision and recall of indel calling. Furthermore, to reliably detect indels, it is essential to choose NGS technologies that offer a long read length and high coverage coupled with specific variant calling tools.     

Haplotype Estimation Using Sequencing Reads

High-throughput sequencing technologies produce short sequence reads that can contain phase information if they span two or more heterozygote genotypes. This information is not routinely used by current methods that infer haplotypes from genotype data. We have extended the SHAPEIT2 method to use phase-informative sequencing reads to improve phasing accuracy. Our model incorporates the read information in a probabilistic model through base quality scores within each read. The method is primarily designed for high-coverage sequence data or data sets that already have genotypes called. One important application is phasing of single samples sequenced at high coverage for use in medical sequencing and studies of rare diseases. Our method can also use existing panels of reference haplotypes. We tested the method by using a mother-father-child trio sequenced at high-coverage by Illumina together with the low-coverage sequence data from the 1000 Genomes Project (1000GP). We found that use of phase-informative reads increases the mean distance between switch errors by 22% from 274.4 kb to 328.6 kb. We also used male chromosome X haplotypes from the 1000GP samples to simulate sequencing reads with varying insert size, read length, and base error rate. When using short 100 bp paired-end reads, we found that using mixtures of insert sizes produced the best results. When using longer reads with high error rates (5–20 kb read with 4%–15% error per base), phasing performance was substantially improved.     

A MinION™‐based pipeline for fast and cost‐effective DNA barcoding

DNA barcodes are useful for species discovery and species identification, but obtaining barcodes currently requires a well‐equipped molecular laboratory and is time‐consuming, and/or expensive. We here address these issues by developing a barcoding pipeline for Oxford Nanopore MinION? and demonstrating that one flow cell can generate barcodes for ~500 specimens despite the high basecall error rates of MinION? reads. The pipeline overcomes these errors by first summarizing all reads for the same tagged amplicon as a consensus barcode. Consensus barcodes are overall mismatch‐free but retain indel errors that are concentrated in homopolymeric regions. They are addressed with an optional error correction pipeline that is based on conserved amino acid motifs from publicly available barcodes. The effectiveness of this pipeline is documented by analysing reads from three MinION? runs that represent three different stages of MinION? development. They generated data for (i) 511 specimens of a mixed Diptera sample, (ii) 575 specimens of ants and (iii) 50 specimens of Chironomidae. The run based on the latest chemistry yielded MinION? barcodes for 490 of the 511 specimens which were assessed against reference Sanger barcodes (N = 471). Overall, the MinION? barcodes have an accuracy of 99.3%–100% with the number of ambiguous bases after correction ranging from <0.01% to 1.5% depending on which correction pipeline is used. We demonstrate that it requires ~2 hr of sequencing to gather all information needed for obtaining reliable barcodes for most specimens (>90%). We estimate that up to 1,000 barcodes can be generated in one flow cell and that the cost per barcode can be 相似文献