CoChat
Sign Up
Cited by 109

The DOI matches its Crossref record; 22 of the 22 references it cites were found in a registry or an index.

De novo detection of somatic mutations in high-throughput single-cell profiling data sets

Muyas, Francesc, Sauer, Carolin M., Valle-Inclán, Jose Espejo, Li, Ruoyan, Rahbari, Raheleh, Mitchell, Thomas J., Hormoz, Sahand, Cortés-Ciriano, Isidro·2024·Nature Biotechnology
DOI 10.1038/s41587-023-01863-zPMC PMC11098751

Published version of a preprint

Open this paper in CoChat — free account

Abstract

Characterization of somatic mutations at single-cell resolution is essential to study cancer evolution, clonal mosaicism and cell plasticity. Here, we describe SComatic, an algorithm designed for the detection of somatic mutations in single-cell transcriptomic and ATAC-seq (assay for transposase-accessible chromatin sequence) data sets directly without requiring matched bulk or single-cell DNA sequencing data. SComatic distinguishes somatic mutations from polymorphisms, RNA-editing events and artefacts using filters and statistical tests parameterized on non-neoplastic samples. Using >2.6 million single cells from 688 single-cell RNA-seq (scRNA-seq) and single-cell ATAC-seq (scATAC-seq) data sets spanning cancer and non-neoplastic samples, we show that SComatic detects mutations in single cells accurately, even in differentiated cells from polyclonal tissues that are not amenable to mutation detection using existing methods. Validated against matched genome sequencing and scRNA-seq data, SComatic achieves F1 scores between 0.6 and 0.7 across diverse data sets, in comparison to 0.2–0.4 for the second-best performing method. In summary, SComatic permits de novo mutational signature analysis, and the study of clonal heterogeneity and mutational burdens at single-cell resolution.

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 · 10.1038/s41587-023-01863-z

Cites

22 of 22 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1 Overview of SComatic. Methodology for detecting somatic mutations in high-throughput single-cell profiling data sets. The dashed red line shows an arbitrarily chosen level of significance for illustration purposes.
Fig. 1 Overview of SComatic. Methodology for detecting somatic mutations in high-throughput single-cell profiling data sets. The dashed red line shows an arbitrarily chosen level of significance for illustration purposes.CC BY · PubMed Central
Fig. 2 Validation of SComatic using matched scRNA-seq and exome sequencing data. a, Mutational burdens for epithelial cells using the somatic SNVs detected by SComatic in cSCC and matched normal skin scRNA-seq data sets. The number of mutations is normalized to account for the variable number of callable sites in each sample. b, Fraction of somatic SNVs detected in epithelial cells attributed to COSMIC signatures. SBS signatures associated with ultraviolet radiation (SBS7a, SBS7b, SBS7c and SBS7d) and clock-like mutational processes (SBS5 and SBS40) are collapsed for visualization purposes. c, Mutational spectra computed for the mutations detected using SComatic in epithelial cells from cSCC and matched normal skin scRNA-seq data. The cosine similarities between the observed and reconstructed mutational spectra are shown. d, Venn diagram showing the overlap of the somatic SNVs detected by SComatic in epithelial cells using scRNA-seq data and WES data from the cSCC samples. ‘WES-specific beta-binomial’ refers to mutations detected in WES with at least one alternative read count in scRNA-seq that are not significant for the beta-binomial test. e, Decomposition of the mutations detected in scRNA-seq data only (scRNA-seq-specific mutations) into COSMIC signatures. f, Correlation between the mutational burdens estimated using the mutations detected in WES and the mutations detected by SComatic in the scRNA-seq data. The correlation was assessed using a linear regression model. Only genomic regions with sufficient sequencing depth in both the WES and scRNA-seq data were considered for this analysis. Mb, megabase. Source data
Fig. 2 Validation of SComatic using matched scRNA-seq and exome sequencing data. a, Mutational burdens for epithelial cells using the somatic SNVs detected by SComatic in cSCC and matched normal skin scRNA-seq data sets. The number of mutations is normalized to account for the variable number of callable sites in each sample. b, Fraction of somatic SNVs detected in epithelial cells attributed to COSMIC signatures. SBS signatures associated with ultraviolet radiation (SBS7a, SBS7b, SBS7c and SBS7d) and clock-like mutational processes (SBS5 and SBS40) are collapsed for visualization purposes. c, Mutational spectra computed for the mutations detected using SComatic in epithelial cells from cSCC and matched normal skin scRNA-seq data. The cosine similarities between the observed and reconstructed mutational spectra are shown. d, Venn diagram showing the overlap of the somatic SNVs detected by SComatic in epithelial cells using scRNA-seq data and WES data from the cSCC samples. ‘WES-specific beta-binomial’ refers to mutations detected in WES with at least one alternative read count in scRNA-seq that are not significant for the beta-binomial test. e, Decomposition of the mutations detected in scRNA-seq data only (scRNA-seq-specific mutations) into COSMIC signatures. f, Correlation between the mutational burdens estimated using the mutations detected in WES and the mutations detected by SComatic in the scRNA-seq data. The correlation was assessed using a linear regression model. Only genomic regions with sufficient sequencing depth in both the WES and scRNA-seq data were considered for this analysis. Mb, megabase. Source dataCC BY · PubMed Central
Fig. 3 Comparison of the performance of SComatic against other mutation detection methods. a–c, Performance of Strelka2, SAMtools, VarScan2, Monovar, SCReadCounts and SComatic for the detection of somatic mutations in the scRNA-seq data from cSCC (a), ovarian cancer (b) and kidney tumor samples (c). The bars represent the mean value, and the error bars are the 95% bootstrap confidence interval for each statistic computed using 50 bootstrap resamples. Significance with respect to SComatic in a–c was assessed using the two-sided Student’s t-test (***P < 0.0001). d, Decomposition into COSMIC signatures of the mutations detected in cSCC scRNA-seq data and in matched WES data. e, Decomposition into COSMIC signatures of the mutations detected in scRNA-seq and matched WGS data from ovarian cancer samples. f, Decomposition into COSMIC signatures of the mutations detected by SComatic in scRNA-seq from homologous recombination deficient (HRD) and homologous recombination proficient (HRP) ovarian cancer samples. g, Comparison between the mutational spectra of the mutations detected in cSCC samples using WES and scRNA-seq data for the algorithms benchmarked. The cosine similarities between the mutational spectra computed using the mutations detected in the scRNA-seq and the WES data are shown. Source data
Fig. 3 Comparison of the performance of SComatic against other mutation detection methods. a–c, Performance of Strelka2, SAMtools, VarScan2, Monovar, SCReadCounts and SComatic for the detection of somatic mutations in the scRNA-seq data from cSCC (a), ovarian cancer (b) and kidney tumor samples (c). The bars represent the mean value, and the error bars are the 95% bootstrap confidence interval for each statistic computed using 50 bootstrap resamples. Significance with respect to SComatic in a–c was assessed using the two-sided Student’s t-test (***P < 0.0001). d, Decomposition into COSMIC signatures of the mutations detected in cSCC scRNA-seq data and in matched WES data. e, Decomposition into COSMIC signatures of the mutations detected in scRNA-seq and matched WGS data from ovarian cancer samples. f, Decomposition into COSMIC signatures of the mutations detected by SComatic in scRNA-seq from homologous recombination deficient (HRD) and homologous recombination proficient (HRP) ovarian cancer samples. g, Comparison between the mutational spectra of the mutations detected in cSCC samples using WES and scRNA-seq data for the algorithms benchmarked. The cosine similarities between the mutational spectra computed using the mutations detected in the scRNA-seq and the WES data are shown. Source dataCC BY · PubMed Central
Fig. 4 Detection of somatic mutations in scRNA-seq data from colorectal cancer samples. a, Mutational burden of epithelial cells computed using SComatic. The number of mutations is normalized to the number of callable sites per sample. b, Distribution of the mutational burden of epithelial cells from MSI tumors detected using SComatic and the mutational burden of MSI tumors from TCGA computed using WES data. The red horizontal line shows the mean for each group, and n indicates the number of samples per group. Statistical significance was assessed using the two-sided Student’s t-test. c, Decomposition of the mutational spectra computed using SComatic into COSMIC signatures. Mutational signatures associated with MMR deficiency (MMRd) (SBS6, SBS14, SBS15, SBS21, SBS26 and SBS44), POLE deficiency (POLEd) (SBS10a, SBS10b and SBS28) and clock-like mutational processes (SBS5 and SBS40) are collapsed for visualization purposes. d, Trinucleotide context of somatic mutations detected by SComatic using the scRNA-seq data from colorectal cancer samples. CRC, colorectal cancer; TCGA, The Cancer Genome Atlas. Source data
Fig. 4 Detection of somatic mutations in scRNA-seq data from colorectal cancer samples. a, Mutational burden of epithelial cells computed using SComatic. The number of mutations is normalized to the number of callable sites per sample. b, Distribution of the mutational burden of epithelial cells from MSI tumors detected using SComatic and the mutational burden of MSI tumors from TCGA computed using WES data. The red horizontal line shows the mean for each group, and n indicates the number of samples per group. Statistical significance was assessed using the two-sided Student’s t-test. c, Decomposition of the mutational spectra computed using SComatic into COSMIC signatures. Mutational signatures associated with MMR deficiency (MMRd) (SBS6, SBS14, SBS15, SBS21, SBS26 and SBS44), POLE deficiency (POLEd) (SBS10a, SBS10b and SBS28) and clock-like mutational processes (SBS5 and SBS40) are collapsed for visualization purposes. d, Trinucleotide context of somatic mutations detected by SComatic using the scRNA-seq data from colorectal cancer samples. CRC, colorectal cancer; TCGA, The Cancer Genome Atlas. Source dataCC BY · PubMed Central
Fig. 5 Detection of somatic mutations in samples with a low tumor mutational burden. a, Trinucleotide context of somatic mutations detected in HSCs from patients with MPNs. b, Decomposition of the somatic mutations detected in HSCs from patients with MPNs into COSMIC signatures. c, Correlation between the mutational burden of HSCs estimated using SComatic and the age of patients at the time of sampling (Pearson’s correlation test). d, Average number of mutations detected per cell and genome in cardiomyocytes from the heart cell atlas across donors. e, Decomposition of the mutations detected in cardiomyocytes into COSMIC signatures. f, Trinucleotide context of mutations detected in cardiomyocytes from the heart cell atlas. g, Average mutational burden of individual cells across the tissues included in the GTEx scRNA-seq data set. h, Decomposition of the mutations detected across all cells from the GTEx data set into COSMIC signatures. i, Trinucleotide context of mutations detected across all single cells from the GTEx data set. The numbers on top of the bars in d and g indicate the number of cells per cell type analyzed, and the horizontal red dashed line corresponds to 1,000 mutations per cell. Source data
Fig. 5 Detection of somatic mutations in samples with a low tumor mutational burden. a, Trinucleotide context of somatic mutations detected in HSCs from patients with MPNs. b, Decomposition of the somatic mutations detected in HSCs from patients with MPNs into COSMIC signatures. c, Correlation between the mutational burden of HSCs estimated using SComatic and the age of patients at the time of sampling (Pearson’s correlation test). d, Average number of mutations detected per cell and genome in cardiomyocytes from the heart cell atlas across donors. e, Decomposition of the mutations detected in cardiomyocytes into COSMIC signatures. f, Trinucleotide context of mutations detected in cardiomyocytes from the heart cell atlas. g, Average mutational burden of individual cells across the tissues included in the GTEx scRNA-seq data set. h, Decomposition of the mutations detected across all cells from the GTEx data set into COSMIC signatures. i, Trinucleotide context of mutations detected across all single cells from the GTEx data set. The numbers on top of the bars in d and g indicate the number of cells per cell type analyzed, and the horizontal red dashed line corresponds to 1,000 mutations per cell. Source dataCC BY · PubMed Central
Fig. 6 Analysis of intra-tumor heterogeneity using somatic mutations detected by SComatic in the scRNA-seq data from a patient with ovarian cancer (SPECTRUM-OV-003). a, Hierarchical clustering of single cells from all tumor regions (columns) by somatic mutations (rows; mutations are labeled arbitrarily). Mutations detected in the scRNA-seq data are shown in red. White denotes the absence of mutations in the scRNA-seq data in cases when the site was sufficiently covered (at least one sequencing read), and gray indicates that there was no coverage at the position to make a call. b, Hierarchical clustering of single cells collected from the upper right quadrant region from patient SPECTRUM-OV-003. Only the mutations shown in a that were detected in at least 20 cells are shown. The two clones defined by somatic mutations detected in scRNA-seq data are marked on the y axis. Single cells and mutations in a and b are ordered by hierarchical clustering (top and left-hand side dendrograms, respectively). The color bar indicates the cancer cell fraction (CCF) of the mutations in the WGS data. NA, no coverage in scRNA-seq. Source data
Fig. 6 Analysis of intra-tumor heterogeneity using somatic mutations detected by SComatic in the scRNA-seq data from a patient with ovarian cancer (SPECTRUM-OV-003). a, Hierarchical clustering of single cells from all tumor regions (columns) by somatic mutations (rows; mutations are labeled arbitrarily). Mutations detected in the scRNA-seq data are shown in red. White denotes the absence of mutations in the scRNA-seq data in cases when the site was sufficiently covered (at least one sequencing read), and gray indicates that there was no coverage at the position to make a call. b, Hierarchical clustering of single cells collected from the upper right quadrant region from patient SPECTRUM-OV-003. Only the mutations shown in a that were detected in at least 20 cells are shown. The two clones defined by somatic mutations detected in scRNA-seq data are marked on the y axis. Single cells and mutations in a and b are ordered by hierarchical clustering (top and left-hand side dendrograms, respectively). The color bar indicates the cancer cell fraction (CCF) of the mutations in the WGS data. NA, no coverage in scRNA-seq. Source dataCC 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

Research works printed but not linked: 51

  1. Neftel, C. et al. An integrative model of cellular states, plasticity, and genetics for glioblastoma. Cell 178, 835–849.e21 (2019). · DOI · Paper page
  2. Kakiuchi, N. & Ogawa, S. Clonal expansion in non-cancer tissues. Nat. Rev. Cancer 21, 239–256 (2021). · DOI · Paper page
  3. Nam, A. S., Chaligne, R. & Landau, D. A. Integrating genetic and non-genetic determinants of cancer evolution by single-cell multi-omics. Nat. Rev. Genet. 22, 3–18 (2021). · DOI
  4. Lim, B., Lin, Y. & Navin, N. Advancing cancer research and medicine with single-cell genomics. Cancer Cell 37, 456–470 (2020). · DOI
  5. Gawad, C., Koh, W. & Quake, S. R. Single-cell genome sequencing: current state of the science. Nat. Rev. Genet. 17, 175–188 (2016). · DOI · Paper page
  6. Xing, D., Tan, L., Chang, C.-H., Li, H. & Xie, X. S. Accurate SNV detection in single cells by transposon-based whole-genome amplification of complementary strands. Proc. Natl Acad. Sci. USA 118, e2013106118 (2021). · DOI
  7. Abascal, F. et al. Somatic mutation landscapes at single-molecule resolution. Nature 593, 405–410 (2021). · DOI · Paper page
  8. Li, R. et al. Mapping single-cell transcriptomes in the intra-tumoral and associated territories of kidney cancer. Cancer Cell 40, 1583–1599.e10 (2022). · DOI
  9. Macaulay, I. C. et al. G&T-seq: parallel sequencing of single-cell genomes and transcriptomes. Nat. Methods 12, 519–522 (2015). · DOI · Paper page
  10. Reuter, J. A., Spacek, D. V., Pai, R. K. & Snyder, M. P. Simul-seq: combined DNA and RNA sequencing for whole-genome and transcriptome profiling. Nat. Methods 13, 953–958 (2016). · DOI
  11. Fan, J. et al. Linking transcriptional and genetic tumor heterogeneity through allele analysis of single-cell RNA-seq data. Genome Res. 28, 1217–1227 (2018). · DOI
  12. Kharchenko, P. V., Silberstein, L. & Scadden, D. T. Bayesian approach to single-cell differential expression analysis. Nat. Methods 11, 740–742 (2014). · DOI · Paper page
  13. Huang, A. Y. et al. Parallel RNA and DNA analysis after deep sequencing (PRDD-seq) reveals cell type-specific lineage patterns in human brain. Proc. Natl Acad. Sci. USA 117, 13886–13895 (2020). · DOI
  14. McCarthy, D. J. et al. Cardelino: computational integration of somatic clonal substructure and single-cell transcriptomes. Nat. Methods 17, 414–421 (2020). · DOI
  15. Liu, F. et al. Systematic comparative analysis of single-nucleotide variant detection methods from single-cell RNA sequencing data. Genome Biol. 20, 242 (2019). · DOI
  16. Coorens, T. H. H. et al. Extensive phylogenies of human development inferred from somatic mutations. Nature 597, 387–392 (2021). · DOI
  17. Karczewski, K. J. et al. The mutational constraint spectrum quantified from variation in 141,456 humans. Nature 581, 434–443 (2020). · DOI · Paper page
  18. Reble, E., Castellani, C. A., Melka, M. G., O’Reilly, R. & Singh, S. M. VarScan2 analysis of de novo variants in monozygotic twins discordant for schizophrenia. Psychiatr. Genet. 27, 62–70 (2017). · DOI
  19. Li, H. & Durbin, R. Fast and accurate short read alignment with Burrows–Wheeler transform. Bioinformatics 25, 1754–1760 (2009). · DOI · Paper page
  20. Kim, S. et al. Strelka2: fast and accurate calling of germline and somatic variants. Nat. Methods 15, 591–594 (2018). · DOI
  21. Zafar, H., Wang, Y., Nakhleh, L., Navin, N. & Chen, K. Monovar: single-nucleotide variant detection in single cells. Nat. Methods 13, 505–507 (2016). · DOI
  22. Prashant, N. M. et al. SCReadCounts: estimation of cell-level SNVs expression from scRNA-seq data. BMC Genomics 22, 689 (2021). · DOI
  23. Vázquez-García, I. et al. Ovarian cancer mutational processes drive site-specific immune evasion. Nature 612, 778–786 (2022). · DOI
  24. Li, R. et al. Mapping single-cell transcriptomes in the intra-tumoral and associated territories of kidney cancer. Cancer Cell 40, 1583–1599.e10 (2022). · DOI
  25. Gulhan, D. C., Lee, J. J.-K., Melloni, G. E. M., Cortés-Ciriano, I. & Park, P. J. Detecting the mutational signature of homologous recombination deficiency in clinical samples. Nat. Genet. 51, 912–919 (2019). · DOI
  26. Pelka, K. et al. Spatially organized multicellular immune hubs in human colorectal cancer. Cell 184, 4734–4752.e20 (2021). · DOI
  27. Lee, H.-O. et al. Lineage-dependent gene expression programs influence the immune landscape of colorectal cancer. Nat. Genet. 52, 594–603 (2020). · DOI
  28. Cortes-Ciriano, I., Lee, S., Park, W.-Y., Kim, T.-M. & Park, P. J. A molecular portrait of microsatellite instability across multiple cancers. Nat. Commun. 8, 15180 (2017). · DOI
  29. Haradhvala, N. J. et al. Distinct mutational signatures characterize concurrent loss of polymerase proofreading and mismatch repair. Nat. Commun. 9, 1746 (2018). · DOI
  30. Alexandrov, L. B. et al. The repertoire of mutational signatures in human cancer. Nature 578, 94–101 (2020). · DOI · Paper page
  31. Osorio, F. G. et al. Somatic mutations reveal lineage relationships and age-related mutagenesis in human hematopoiesis. Cell Rep. 25, 2308–2316.e4 (2018). · DOI · Paper page
  32. Choudhury, S. et al. Somatic mutations in single human cardiomyocytes reveal age-associated DNA damage and widespread oxidative genotoxicity. Nat. Aging 2, 714–725 (2022). · DOI
  33. Eraslan, G. et al. Single-nucleus cross-tissue molecular reference maps toward understanding disease gene function. Science 376, eabl4290 (2022). · DOI
  34. Zhang, K. et al. A single-cell atlas of chromatin accessibility in the human genome. Cell 184, 5985–6001.e19 (2021). · DOI · Paper page
  35. Gao, T. et al. Haplotype-aware analysis of somatic copy number variations from single-cell transcriptomes. Nat. Biotechnol. 41, 417–426 (2023). · DOI
  36. Van Egeren, D. et al. Transcriptional differences between JAK2-V617F and wild-type bone marrow cells in patients with myeloproliferative neoplasms. Exp. Hematol. 107, 14–19 (2022). · DOI
  37. Rozenblatt-Rosen, O. et al. The Human Tumor Atlas Network: charting tumor transitions across space and time at single-cell resolution. Cell 181, 236–249 (2020). · DOI
  38. Li, H. Aligning sequence reads, clone sequences and assembly contigs with BWA-MEM. Preprint at arXiv https://doi.org/10.48550/arXiv.1303.3997 (2013). · DOI
  39. Muyas, F., Zapata, L., Guigó, R. & Ossowski, S. The rate and spectrum of mosaic mutations during embryogenesis revealed by RNA sequencing of 49 tissues. Genome Med. 12, 49 (2020). · DOI
  40. Bonfield, J. K. et al. HTSlib: C library for reading/writing high-throughput sequencing data. Gigascience 10, giab007 (2021).https://github.com/pysam-developers/pysam · DOI
  41. Lo Giudice, C., Tangaro, M. A., Pesole, G. & Picardi, E. Investigating RNA editing in deep transcriptome datasets with REDItools and REDIportal. Nat. Protoc. 15, 1098–1131 (2020). · DOI
  42. Kiran, A. & Baranov, P. V. DARNED: a DAtabase of RNa EDiting in humans. Bioinformatics 26, 1772–1776 (2010). · DOI
  43. Nakamura, K. et al. Sequence-specific error profile of Illumina sequencers. Nucleic Acids Res. 39, e90 (2011). · DOI
  44. DePristo, M. A. et al. A framework for variation discovery and genotyping using next-generation DNA sequencing data. Nat. Genet. 43, 491–498 (2011). · DOI
  45. Fan, Y. et al. MuSE: accounting for tumor heterogeneity using a sample-specific error model improves sensitivity and specificity in mutation calling from sequencing data. Genome Biol. 17, 178 (2016). · DOI
  46. Dentro, S. C. et al. Characterizing genetic intra-tumor heterogeneity across 2,658 human cancer genomes. Cell 184, 2239–2254.e39 (2021). · DOI · Paper page
  47. Li, H. et al. The sequence alignment/map format and SAMtools. Bioinformatics 25, 2078–2079 (2009). · DOI · Paper page
  48. Koboldt, D. C. et al. VarScan 2: somatic mutation and copy number alteration discovery in cancer by exome sequencing. Genome Res. 22, 568–576 (2012). · DOI
  49. Karczewski, K. J. et al. The ExAC browser: displaying reference data information from over 60 000 exomes. Nucleic Acids Res. 45, D840–D845 (2017). · DOI
  50. Loh, P.-R. et al. Reference-based phasing using the Haplotype Reference Consortium panel. Nat. Genet. 48, 1443–1448 (2016). · DOI · Paper page
  51. Wang, K., Li, M. & Hakonarson, H. ANNOVAR: functional annotation of genetic variants from high-throughput sequencing data. Nucleic Acids Res. 38, e164 (2010). · DOI

Cited by

See all 109

Metadata

Authors
Muyas, F., Sauer, C. M., Valle-Inclán, J. E., Li, R., Rahbari, R., Mitchell, T. J., Hormoz, S., & Cortés-Ciriano, I.
Venue
Nature Biotechnology
Publisher
Springer Science and Business Media LLC
Published
2023-07-06
Volume
42
Issue
5
Pages
758-767
ISSN
  • 1087-0156
  • 1546-1696
DOI
10.1038/s41587-023-01863-z
PMC
PMC11098751
Cited by
109
References
22
Language
en
Licence
Funders
  • European Molecular Biology Laboratory
  • Wellcome Trust
  • Harvard Medical School

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