CoChat
Sign Up
Cited by 438Genomics and Phylogenetic StudiesGenetic Associations and EpidemiologyGenetic diversity and population structure

The DOI matches its Crossref record; 40 of the 40 references it cites were found in a registry or an index.

SNP Calling, Genotype Calling, and Sample Allele Frequency Estimation from New-Generation Sequencing Data

Open access
Nielsen, Rasmus, Korneliussen, Thorfinn, Albrechtsen, Anders, Li, Yingrui, Wang, Jun·2012·PLoS ONE
DOI 10.1371/journal.pone.0037558PMC PMC3404070
Open this paper in CoChat — free account

Abstract

We present a statistical framework for estimation and application of sample allele frequency spectra from New-Generation Sequencing (NGS) data. In this method, we first estimate the allele frequency spectrum using maximum likelihood. In contrast to previous methods, the likelihood function is calculated using a dynamic programming algorithm and numerically optimized using analytical derivatives. We then use a bayesian method for estimating the sample allele frequency in a single site, and show how the method can be used for genotype calling and SNP calling. We also show how the method can be extended to various other cases including cases with deviations from Hardy-Weinberg equilibrium. We evaluate the statistical properties of the methods using simulations and by application to a real data set.

Overview

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Integrity signals

Registry

Crossref · 10.1371/journal.pone.0037558

Cites

40 of 40 references verified0 citation errors · 0 not checkable

Statistics

Figures

Figure 1 The distribution of true (True) and estimated unfolded SFS using the Maximum Likelihood method (ML) presented in the paper, genotype calling based on choosing the genotype with highest posterior probability (GC), and using the Bayesian procedure described in the text (Bay) in a sample from 50 MB 10 diploid individuals, where 2% of all SNPs are variable in the population and follow a distribution of allele frequencies, p, proportional to 1/p. An error rate of 0.5% is assumed. The mean sequencing depths are 1X (a), 3X (b), 5X (c), and 10X (d). The values presented in the figure legend box are the estimates of the proportion of sites that are variable in the sample.
Figure 1 The distribution of true (True) and estimated unfolded SFS using the Maximum Likelihood method (ML) presented in the paper, genotype calling based on choosing the genotype with highest posterior probability (GC), and using the Bayesian procedure described in the text (Bay) in a sample from 50 MB 10 diploid individuals, where 2% of all SNPs are variable in the population and follow a distribution of allele frequencies, p, proportional to 1/p. An error rate of 0.5% is assumed. The mean sequencing depths are 1X (a), 3X (b), 5X (c), and 10X (d). The values presented in the figure legend box are the estimates of the proportion of sites that are variable in the sample.CC BY · PubMed Central
Figure 2 ROC curves for different SNP callers. Data for 10 individuals were simulated assuming a sequencing depth of 2 and a raw sequencing error rate of 1% (A) and (B) a depth of 5 and a raw sequencing error rate of 5%. The SFS method is the main method described in the text. The GC method is based on genotype calling using the genotype with the highest posterior probability. The LR method is based on a likelihood ratio test of the hypothesis that the allele frequency is zero. The SFS based method and the LR method have similar performance except for very high error rates, where the SFS tends to be somewhat better. Both methods in general perform much better than the GC method. The difference would even larger in larger panels of individuals. Simulations under other conditions can be found in Figure S1.
Figure 2 ROC curves for different SNP callers. Data for 10 individuals were simulated assuming a sequencing depth of 2 and a raw sequencing error rate of 1% (A) and (B) a depth of 5 and a raw sequencing error rate of 5%. The SFS method is the main method described in the text. The GC method is based on genotype calling using the genotype with the highest posterior probability. The LR method is based on a likelihood ratio test of the hypothesis that the allele frequency is zero. The SFS based method and the LR method have similar performance except for very high error rates, where the SFS tends to be somewhat better. Both methods in general perform much better than the GC method. The difference would even larger in larger panels of individuals. Simulations under other conditions can be found in Figure S1.CC BY · PubMed Central
Figure 3 The error rate of different genotype callers for different call rates. The SFS-method is the method described in the main text. The MAF method is based on first obtaining a maximum likelihood estimate of the allele frequency, and then use the estimated allele frequency to define priors for genotype calling. The GC-max method is based on calling genotypes with highest posterior probability. The GC-ratio method is based on calling genotypes depending on the ratio of the likelihood for the most likely to second most likely genotype. The jagged behavior of some of the curves is a consequence of the discrete nature of the data, i.e. an individual contains a discrete number of copies of the minor allele. 10 individuals are simulated for 50,000 variable sites with a distribution of allele frequencies (p), proportional to 1/p with an error rate of 0.5%. Results for other error rates are shown in Figure S2.
Figure 3 The error rate of different genotype callers for different call rates. The SFS-method is the method described in the main text. The MAF method is based on first obtaining a maximum likelihood estimate of the allele frequency, and then use the estimated allele frequency to define priors for genotype calling. The GC-max method is based on calling genotypes with highest posterior probability. The GC-ratio method is based on calling genotypes depending on the ratio of the likelihood for the most likely to second most likely genotype. The jagged behavior of some of the curves is a consequence of the discrete nature of the data, i.e. an individual contains a discrete number of copies of the minor allele. 10 individuals are simulated for 50,000 variable sites with a distribution of allele frequencies (p), proportional to 1/p with an error rate of 0.5%. Results for other error rates are shown in Figure S2.CC BY · PubMed Central
Figure 4 The unfolded site frequency spectrum from 25 Danish indivuduals. The data were previously analyzed in Yi et al. 2010.
Figure 4 The unfolded site frequency spectrum from 25 Danish indivuduals. The data were previously analyzed in Yi et al. 2010.CC BY · PubMed Central

Verification

Claims the paper makes

The claims and statistics of this paper haven’t been checked yet.

Run the check — free, no account needed. It reads the open-access text and checks the sources it cites.

A few free checks a day · papers already verified open instantly · results are public

Claims other papers make about this one

Each is checked against this paper’s own text: does it say what the citing paper says it does?

Self-check

Percentages, sums and reported differences re-computed from the paper’s own numbers.

The self-check needs the paper’s full text.

Citations

Cites

Show all 39 references

Cited by

Sorted by citations

See all 438

Related

Metadata

Authors
Nielsen, R., Korneliussen, T., Albrechtsen, A., Li, Y., & Wang, J.
Affiliations
  • BGI Group (China) (CN)
  • University of Copenhagen (DK)
Corresponding author
  • Nielsen, Rasmus
  • Wang, Jun
ORCID
Venue
PLoS ONE
Publisher
Public Library of Science (PLoS)
Published
2012-07-24
Volume
7
Issue
7
Pages
e37558
ISSN
1932-6203 (ISSN-L)
DOI
10.1371/journal.pone.0037558
PMC
PMC3404070
OpenAlex
W2005287518
Cited by
424
References
39
Citations by year
2026: 24 · 2025: 18 · 2024: 43 · 2023: 33 · 2022: 38 · 2021: 40 · 2020: 25 · 2019: 19 · 2018: 37 · 2017: 30 · 2016: 39 · 2015: 30 · 2014: 25 · 2013: 19 · 2012: 1
FWCI
9.58
Type
journal article
Language
en
Open access
Licence
Locations
  • PLoS ONE (publishedVersion, cc-by)
  • Europe PMC (PubMed Central) (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • Research at the University of Copenhagen (University of Copenhagen) (publishedVersion, cc-by)
  • RePEc: Research Papers in Economics (submittedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion)
  • eScholarship (California Digital Library) (submittedVersion)
  • eScholarship (California Digital Library) (submittedVersion, cc-by)
  • Figshare (submittedVersion, cc-by)
Funders
  • Villum Fonden (00007171)
  • National Institutes of Health
  • National Human Genome Research Institute (R01HG003229)
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Genomics and Phylogenetic Studies
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Genetics › Genetic Associations and Epidemiology
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Genetics › Genetic diversity and population structure
Keywords
allele frequency estimation, allele frequency spectrum, SNP calling, genotype calling, next-generation sequencing, maximum likelihood estimation, Hardy-Weinberg equilibrium, genetic variation
MeSH terms
Algorithms, Bayes Theorem, Computer Simulation, Denmark, Gene Frequency, Genotype, Humans, ROC Curve, Software, Likelihood Functions, Computational Biology, Internet, Polymorphism, Single Nucleotide, White People, High-Throughput Nucleotide Sequencing, Exome

About these checks

Registry, retraction, Cites and statistics results are computed, not judged, with one exception: a printed reference that no index lists is read once by a model before it counts as a citation error. Claims are extracted by a model and kept only when their quote appears verbatim in the text; each cited source is then checked, with a model’s help, for whether it says what the claim relies on. Those verdicts can be wrong — read the claim’s quote and the cited source before relying on one.

Statistics and arithmetic are recomputed, not judged.

Confidence bands are shown only for fields calibrated on our annotated test set. None is calibrated yet, so no band is shown.

What counts as a claim ›How we measure confidence ›How we check this ›

Are you an author, or do you see a problem with this page? Request a review