CoChat
Sign Up
Cited by 109vaccines and immunoinformatics approachesAdvanced Proteomics Techniques and ApplicationsCancer Immunotherapy and Biomarkers

The DOI matches its Crossref record; 78 of the 79 references it cites were found in a registry or an index.

Multiplexed relative and absolute quantitative immunopeptidomics reveals MHC I repertoire alterations induced by CDK4/6 inhibition

Open access
Stopfer, Lauren E., Mesfin, Joshua M., Joughin, Brian A., Lauffenburger, Douglas A., White, Forest M.·2020·Nature Communications
DOI 10.1038/s41467-020-16588-9PMC PMC7265461

Published version of a preprint

Open this paper in CoChat — free accountOpen PDF

Abstract

Peptides bound to class I major histocompatibility complexes (MHC) play a critical role in immune cell recognition and can trigger an antitumor immune response in cancer. Surface MHC levels can be modulated by anticancer agents, altering immunity. However, understanding the peptide repertoire's response to treatment remains challenging and is limited by quantitative mass spectrometry-based strategies lacking normalization controls. We describe an experimental platform that leverages recombinant heavy isotope-coded peptide MHCs (hipMHCs) and multiplex isotope tagging to quantify peptide repertoire alterations using low sample input. HipMHCs improve quantitative accuracy of peptide repertoire changes by normalizing for variation across analyses and enable absolute quantification using internal calibrants to determine copies per cell of MHC antigens, which can inform immunotherapy design. Applying this platform in melanoma cell lines to profile the immunopeptidome response to CDK4/6 inhibition and interferon-γ - known modulators of antigen presentation - uncovers treatment-specific alterations, connecting the intracellular response to extracellular immune presentation.

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/s41467-020-16588-9

Cites

78 of 78 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1 Platform for quantitative immunopeptidomics using hipMHCs. a HipMHCs were generated through UV-mediated peptide exchange of HLA*A2:01 monomers with a heavy leucine HLA-A*02:01 binding peptide. Stable hipMHCs concentrations were measured with an ELISA, and hipMHC complexes were added to lysate samples prior to immunoprecipitation (IP), at the same concentration for quantification correction (blue/teal) or titrated in to create an internal standard curve (red). Heavy and light pMHCs were isolated with IP, acid elution, and molecular weight cut-off (MWCO) filters. b Peptides were analyzed by LC-MS/MS three ways. Relative quantification label-free analyses were quantified by integrating the area under the curve (AUC) of the chromatographic elution across samples, and quantification was normalized by applying correction factors determined by hipMHC AUC intensity ratios between samples. Samples for multiplexed analysis were TMT-labeled and relative quantification was implemented using reporter ion intensities. Normalization was performed using hipMHC reporter ion intensity ratios across TMT channels. For absolute quantification, TMT-labeled samples containing a hipMHC internal standard curve were used to calculate the endogenous copies per cell of the pMHC of interest.
Fig. 1 Platform for quantitative immunopeptidomics using hipMHCs. a HipMHCs were generated through UV-mediated peptide exchange of HLA*A2:01 monomers with a heavy leucine HLA-A*02:01 binding peptide. Stable hipMHCs concentrations were measured with an ELISA, and hipMHC complexes were added to lysate samples prior to immunoprecipitation (IP), at the same concentration for quantification correction (blue/teal) or titrated in to create an internal standard curve (red). Heavy and light pMHCs were isolated with IP, acid elution, and molecular weight cut-off (MWCO) filters. b Peptides were analyzed by LC-MS/MS three ways. Relative quantification label-free analyses were quantified by integrating the area under the curve (AUC) of the chromatographic elution across samples, and quantification was normalized by applying correction factors determined by hipMHC AUC intensity ratios between samples. Samples for multiplexed analysis were TMT-labeled and relative quantification was implemented using reporter ion intensities. Normalization was performed using hipMHC reporter ion intensity ratios across TMT channels. For absolute quantification, TMT-labeled samples containing a hipMHC internal standard curve were used to calculate the endogenous copies per cell of the pMHC of interest.CC BY · PubMed Central
Fig. 2 HipMHCs improve quantitation in LF and TMT-labeled samples. a Experimental design. Five LF (orange) and six TMT (blue) technical replicates of 1 × 107 cells + hipMHCs were used to compare LF and TMT quantification. b Peptide sequence and amount of hipMHC added into each sample. L7 denotes heavy-isotope-labeled leucine (+7). ALNEQIARL7 and SLEEPIGHL7 were used as quantification correction hipMHCs, and SVVESVKFL7 was titrated in across samples. For LF analysis, sample #6 was omitted. c Two thousand three hundred and sixty-nine unique LF peptides were identified across five analyses (black), 1352 of which were quantifiable (gray) via AUC quantification, and 589 quantifiable peptides which were identified in all five analyses (orange). One thousand seven hundred and fifty-four unique peptides were quantified with TMT-labeled analyses combining six TMT fractions (blue). d Ninety-two percent of TMT, 82% of LF, and 5.6% of random peptide 9-mers derived from the proteome (gray) are predicted to bind to an HLA allele in MDA-MB-231 cells with an affinity < 500 nM. e Linear fit of titrated hipMHC peptide for LF (left) and TMT (right). Raw r2 = 0.48 (LF) and r2 = 0.91 (TMT), hipMHC-adjusted (adj) r2 = 0.99 (LF) and r2 = 0.96 (TMT). f Distribution of the log2 fold change (FC) of each peptide’s quantification (x) over the mean (μ) peptide quantification across samples for raw (left) and hipMHC adjusted (right). g Gaussian fit of the frequency distribution of log2(FC) of (x) over (μ) for raw and hipMHC-adjusted LF and TMT samples. In all, 99.7% of variance between peptide quantitation (3× SD) is captured within a 1.65 (raw) and 1.52 (adj) FC from the mean for LF samples, and 1.30 (raw) and 1.23 (adj.) for TMT samples. h Median coefficient of variation (CV) for LF (24.27% raw, 20.99% adj) and TMT (14.00% raw, 7.48% adj). Error bars represent the interquartile range. Source data are provided in the Source data file.
Fig. 2 HipMHCs improve quantitation in LF and TMT-labeled samples. a Experimental design. Five LF (orange) and six TMT (blue) technical replicates of 1 × 107 cells + hipMHCs were used to compare LF and TMT quantification. b Peptide sequence and amount of hipMHC added into each sample. L7 denotes heavy-isotope-labeled leucine (+7). ALNEQIARL7 and SLEEPIGHL7 were used as quantification correction hipMHCs, and SVVESVKFL7 was titrated in across samples. For LF analysis, sample #6 was omitted. c Two thousand three hundred and sixty-nine unique LF peptides were identified across five analyses (black), 1352 of which were quantifiable (gray) via AUC quantification, and 589 quantifiable peptides which were identified in all five analyses (orange). One thousand seven hundred and fifty-four unique peptides were quantified with TMT-labeled analyses combining six TMT fractions (blue). d Ninety-two percent of TMT, 82% of LF, and 5.6% of random peptide 9-mers derived from the proteome (gray) are predicted to bind to an HLA allele in MDA-MB-231 cells with an affinity < 500 nM. e Linear fit of titrated hipMHC peptide for LF (left) and TMT (right). Raw r2 = 0.48 (LF) and r2 = 0.91 (TMT), hipMHC-adjusted (adj) r2 = 0.99 (LF) and r2 = 0.96 (TMT). f Distribution of the log2 fold change (FC) of each peptide’s quantification (x) over the mean (μ) peptide quantification across samples for raw (left) and hipMHC adjusted (right). g Gaussian fit of the frequency distribution of log2(FC) of (x) over (μ) for raw and hipMHC-adjusted LF and TMT samples. In all, 99.7% of variance between peptide quantitation (3× SD) is captured within a 1.65 (raw) and 1.52 (adj) FC from the mean for LF samples, and 1.30 (raw) and 1.23 (adj.) for TMT samples. h Median coefficient of variation (CV) for LF (24.27% raw, 20.99% adj) and TMT (14.00% raw, 7.48% adj). Error bars represent the interquartile range. Source data are provided in the Source data file.CC BY · PubMed Central
Fig. 3 Absolute quantification of pMHCs with hipMHC standards and isobaric labeling. a Peptide intensities for TMT-labeled MDA-MB-231 cells from Fig. 2 were determined by AUC quantification. Percentile of abundance represents a peptide’s rank relative to the most abundant peptide. b Experimental design. Normalization standards along with 5, 15, and 50 fmol of BCAP31 and 0.3, 1, and 3 fmol of DDX5 hipMHCs were added to three biological replicates of 1 × 107 MDA-MB-231 cells, and peptides from MHC complexes were isolated, labeled, and analyzed via LC-MS/MS. c Chromatographic elution profiles for the three TMT reporter ion intensities of the hipMHC standard curve (colored), along with the average (n = 3) TMT reporter ion intensity trace of the endogenous peptide (black). Each MS2 scan is represented as a single point, and elution profiles are fitted with a gaussian distribution (line). d Adjusted (hipMHC normalized) apex intensity versus fmol of hipMHC added creates a standard curve from which the endogenous concentration of antigen is calculated. For both linear fits of BCAP31 and DDX5, r2 > 0.999. The endogenous peptide is presented as the mean value ± SD for n = 3 biological replicates. Source data are provided in the Source data file.
Fig. 3 Absolute quantification of pMHCs with hipMHC standards and isobaric labeling. a Peptide intensities for TMT-labeled MDA-MB-231 cells from Fig. 2 were determined by AUC quantification. Percentile of abundance represents a peptide’s rank relative to the most abundant peptide. b Experimental design. Normalization standards along with 5, 15, and 50 fmol of BCAP31 and 0.3, 1, and 3 fmol of DDX5 hipMHCs were added to three biological replicates of 1 × 107 MDA-MB-231 cells, and peptides from MHC complexes were isolated, labeled, and analyzed via LC-MS/MS. c Chromatographic elution profiles for the three TMT reporter ion intensities of the hipMHC standard curve (colored), along with the average (n = 3) TMT reporter ion intensity trace of the endogenous peptide (black). Each MS2 scan is represented as a single point, and elution profiles are fitted with a gaussian distribution (line). d Adjusted (hipMHC normalized) apex intensity versus fmol of hipMHC added creates a standard curve from which the endogenous concentration of antigen is calculated. For both linear fits of BCAP31 and DDX5, r2 > 0.999. The endogenous peptide is presented as the mean value ± SD for n = 3 biological replicates. Source data are provided in the Source data file.CC BY · PubMed Central
Fig. 4 Palbociclib alters the immunopeptidome in melanoma. a Viability at 72 h after drug treatment; data are represented as a fraction (%) of the DMSO control. Calculated IC50s: SKMEL5 = 12.74 μM, SKMEL28 = 14.62 μM, SKMEL2 = 16.98 μM, and IPC298 = 10.62 μM. Data are presented as mean values ± SD for n = 3 experimental replicates for all cell lines, except SKMEL5 (n = 4). b Experimental setup of TMT-labeled immunopeptidomics experiments in melanoma cell lines. c Flow cytometry measurements of surface HLA expression in SKMEL5 cells. Data are represented as % of maximum signal, and the distributions are representative of three independent experiments. d Histogram distribution of log2 fold change (FC) of (palbociclib/DMSO) for unique pMHCs, where FC is calculated from the mean intensity of n = 3 biological replicates per condition. Data are represented as a % of total unique peptides identified. e Volcano plot displaying log2(FC) of 1 μM treated SKMEL5 cells versus significance (mean-adjusted p value, unpaired two-sided t test). Colored points (p < 0.05, log2(FC) > 1.56) correspond to processes in f. f Log2(FC) of significantly enriched peptides from GO term enrichment processes labeled with source protein name. False discovery rate (FDR)-adj. p value < 0.05. g Four-way Venn diagram of the number of source proteins of peptides significantly enriched (top value) or significantly increasing (bottom value) with 1 μM palbociclib. h Log2(FC) of pIRS2 peptide following 1 μM (gray) or 10 μM (black) palbociclib, *p < 0.05, unpaired two-sided t test. i Source protein name, peptide sequence, and log2(FC) of TAAs in SKMEL5 (left) and IPC298 (right) cells. Exact significance values and other source data are reported in the Source Data File.
Fig. 4 Palbociclib alters the immunopeptidome in melanoma. a Viability at 72 h after drug treatment; data are represented as a fraction (%) of the DMSO control. Calculated IC50s: SKMEL5 = 12.74 μM, SKMEL28 = 14.62 μM, SKMEL2 = 16.98 μM, and IPC298 = 10.62 μM. Data are presented as mean values ± SD for n = 3 experimental replicates for all cell lines, except SKMEL5 (n = 4). b Experimental setup of TMT-labeled immunopeptidomics experiments in melanoma cell lines. c Flow cytometry measurements of surface HLA expression in SKMEL5 cells. Data are represented as % of maximum signal, and the distributions are representative of three independent experiments. d Histogram distribution of log2 fold change (FC) of (palbociclib/DMSO) for unique pMHCs, where FC is calculated from the mean intensity of n = 3 biological replicates per condition. Data are represented as a % of total unique peptides identified. e Volcano plot displaying log2(FC) of 1 μM treated SKMEL5 cells versus significance (mean-adjusted p value, unpaired two-sided t test). Colored points (p < 0.05, log2(FC) > 1.56) correspond to processes in f. f Log2(FC) of significantly enriched peptides from GO term enrichment processes labeled with source protein name. False discovery rate (FDR)-adj. p value < 0.05. g Four-way Venn diagram of the number of source proteins of peptides significantly enriched (top value) or significantly increasing (bottom value) with 1 μM palbociclib. h Log2(FC) of pIRS2 peptide following 1 μM (gray) or 10 μM (black) palbociclib, *p < 0.05, unpaired two-sided t test. i Source protein name, peptide sequence, and log2(FC) of TAAs in SKMEL5 (left) and IPC298 (right) cells. Exact significance values and other source data are reported in the Source Data File.CC BY · PubMed Central
Fig. 5 Pathway analysis of palbociclib-altered immunopeptidome. a Normalized enrichment score (NES) of significantly enriched pathways with 10 μM palbociclib, where +/− NES scores reflect enrichment directionality. For all, q < 0.25, and *p < 0.05, **p < 0.01, and ***p < 0.001. b, c String network of protein–protein interactions of all source proteins from E2F peptides (b) significantly decreasing with 10 μM palbociclib, and OxPhos peptides (c) significantly increasing with 1 μM palbociclib for all cell lines, except SKMEL28, where peptides from 10 μM are depicted. Node color corresponds to cell line. d Quantification (n = 9) of MitoTraker green intensity normalized to cell number following 72 h palbociclib treatment. Data are represented as a box and whiskers plot, with whiskers displaying minimum and maximum signal. Significance was determined using Dunnett’s multiple comparisons test for each condition versus DMSO. *p < 0.05 and ****p < 0.0001. e Correlation between log2 fold change (FC) of (palbociclib/DMSO) for RNA expression (y-axis) and pMHC presentation (x-axis) of SKMEL5 cells treated for 72 h with 1 μM palbociclib, r2 = 0.04. FC is calculated from the mean intensity of n = 3 biological replicates per condition. f Significantly enriched pathways using RNA-seq data (p < 0.05, q < 0.25). Annotated pathways reflect pathways also identified in the immunopeptidome analysis (blue), and those that match with previous reported data (red)7. g Log2(FC) for SKMEL5 OxPhos peptides significantly increasing (p < 0.05, blue) with 1 μM palbociclib, and matched log2(FC) of RNA expression (black). Significant differences in RNA expression (palbociclib versus DMSO) are indicated. **p < 0.01 and ****p < 0.0001 (Wald test, Benjamini–Hochberg (BH) adjusted). Exact significance values and other Source data are reported in the Source Data File.
Fig. 5 Pathway analysis of palbociclib-altered immunopeptidome. a Normalized enrichment score (NES) of significantly enriched pathways with 10 μM palbociclib, where +/− NES scores reflect enrichment directionality. For all, q < 0.25, and *p < 0.05, **p < 0.01, and ***p < 0.001. b, c String network of protein–protein interactions of all source proteins from E2F peptides (b) significantly decreasing with 10 μM palbociclib, and OxPhos peptides (c) significantly increasing with 1 μM palbociclib for all cell lines, except SKMEL28, where peptides from 10 μM are depicted. Node color corresponds to cell line. d Quantification (n = 9) of MitoTraker green intensity normalized to cell number following 72 h palbociclib treatment. Data are represented as a box and whiskers plot, with whiskers displaying minimum and maximum signal. Significance was determined using Dunnett’s multiple comparisons test for each condition versus DMSO. *p < 0.05 and ****p < 0.0001. e Correlation between log2 fold change (FC) of (palbociclib/DMSO) for RNA expression (y-axis) and pMHC presentation (x-axis) of SKMEL5 cells treated for 72 h with 1 μM palbociclib, r2 = 0.04. FC is calculated from the mean intensity of n = 3 biological replicates per condition. f Significantly enriched pathways using RNA-seq data (p < 0.05, q < 0.25). Annotated pathways reflect pathways also identified in the immunopeptidome analysis (blue), and those that match with previous reported data (red)7. g Log2(FC) for SKMEL5 OxPhos peptides significantly increasing (p < 0.05, blue) with 1 μM palbociclib, and matched log2(FC) of RNA expression (black). Significant differences in RNA expression (palbociclib versus DMSO) are indicated. **p < 0.01 and ****p < 0.0001 (Wald test, Benjamini–Hochberg (BH) adjusted). Exact significance values and other Source data are reported in the Source Data File.CC BY · PubMed Central
Fig. 6 Quantifying the pMHC repertoire response to IFN-γ stimulation. a RNA-seq (black) and pMHC (blue) log2 fold change (FC) of 1 μM palbociclib/DMSO, calculated from the mean intensity of n = 3 biological replicates per condition, for antigen-processing genes. b Surface HLA expression via flow cytometry of cells treated with 72 h IFN-γ shown as log2(FC) (IFN-γ/DMSO). Errors bars represent ± SD, biological replicates are n = 8, 11, 9, and 9 for SKMEL5, SKMEL28, SKMEL2, and IPC298, respectively. c Immunopeptidome log2(FC), dotted lines display quartiles, and mean fold changes (solid line) are 2.42, 2.50, 2.08, and 3.04 for SKMEL5, SKMEL28, SKMEL2, and IPC298 cells, respectively. d Significantly enriched pathways in SKMEL5 cells with 72 h IFN-γ, q < 0.25, *p < 0.05, **p < 0.01. e Enrichment plot of IFN-γ response enrichment in SKMEL5 cells displays running enrichment score (green, right y-axis), and the log2(FC) (left y-axis) versus rank (x-axis) for each peptide (gray). Open circles show significantly enriched IFN-γ peptides. f Volcano plot of IFN-γ-induced changes in SKMEL5 cells. Peptides are presented as the log2(FC) versus mean-adjusted p value (unpaired two-sided t test). Red points represent peptides significantly enriched (p < 0.05, fold change > 2.42). g Venn diagram of significantly enriched source proteins (black) or peptides (gray) between IFN-γ and 1 μM palbociclib-treated SKMEL5 cells. h Protein–protein interaction network of significantly enriched source proteins in common, annotated by enriched gene ontology cellular components (CCs) and/or biological professes (BPs). Significance values are false discovery rate (FDR) adjusted. Exact significance values and other Source data are reported in the Source Data File.
Fig. 6 Quantifying the pMHC repertoire response to IFN-γ stimulation. a RNA-seq (black) and pMHC (blue) log2 fold change (FC) of 1 μM palbociclib/DMSO, calculated from the mean intensity of n = 3 biological replicates per condition, for antigen-processing genes. b Surface HLA expression via flow cytometry of cells treated with 72 h IFN-γ shown as log2(FC) (IFN-γ/DMSO). Errors bars represent ± SD, biological replicates are n = 8, 11, 9, and 9 for SKMEL5, SKMEL28, SKMEL2, and IPC298, respectively. c Immunopeptidome log2(FC), dotted lines display quartiles, and mean fold changes (solid line) are 2.42, 2.50, 2.08, and 3.04 for SKMEL5, SKMEL28, SKMEL2, and IPC298 cells, respectively. d Significantly enriched pathways in SKMEL5 cells with 72 h IFN-γ, q < 0.25, *p < 0.05, **p < 0.01. e Enrichment plot of IFN-γ response enrichment in SKMEL5 cells displays running enrichment score (green, right y-axis), and the log2(FC) (left y-axis) versus rank (x-axis) for each peptide (gray). Open circles show significantly enriched IFN-γ peptides. f Volcano plot of IFN-γ-induced changes in SKMEL5 cells. Peptides are presented as the log2(FC) versus mean-adjusted p value (unpaired two-sided t test). Red points represent peptides significantly enriched (p < 0.05, fold change > 2.42). g Venn diagram of significantly enriched source proteins (black) or peptides (gray) between IFN-γ and 1 μM palbociclib-treated SKMEL5 cells. h Protein–protein interaction network of significantly enriched source proteins in common, annotated by enriched gene ontology cellular components (CCs) and/or biological professes (BPs). Significance values are false discovery rate (FDR) adjusted. Exact significance values and other Source data are reported in the Source Data File.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 79 references

Cited by

Sorted by citations

Related

See all

Metadata

Authors
Stopfer, L. E., Mesfin, J. M., Joughin, B. A., Lauffenburger, D. A., & White, F. M.
Affiliations
  • Koch Institute for Integrative Cancer Research At MIT (US)
  • Massachusetts Institute of Technology (US)
Corresponding author
White, Forest M.
ORCID
Venue
Nature Communications
Publisher
Springer Science and Business Media LLC
Published
2020-06-02
Volume
11
Issue
1
Pages
2760-2760
ISSN
2041-1723 (ISSN-L)
DOI
10.1038/s41467-020-16588-9
PMC
PMC7265461
OpenAlex
W3031716951
Cited by
109
References
79
Citations by year
2026: 17 · 2025: 21 · 2024: 11 · 2023: 23 · 2022: 14 · 2021: 16 · 2020: 6
FWCI
4.76
Type
journal article
Language
en
Open access
Licence
Locations
  • Nature Communications (publishedVersion, cc-by)
  • bioRxiv (Cold Spring Harbor Laboratory) (acceptedVersion)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • RePEc: Research Papers in Economics (submittedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • DSpace@MIT (Massachusetts Institute of Technology) (submittedVersion, cc-by-nc)
Funders
  • Melanoma Research Alliance (565436)
  • Array BioPharma
  • Koch Institute for Integrative Cancer Research, Massachusetts Institute of Technology
  • National Institutes of Health (CA210180)
  • National Heart, Lung, and Blood Institute (T32HL105373)
  • National Cancer Institute (U54CA210180, P30CA014051)
  • National Institute of Environmental Health Sciences (T32ES007020)
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › vaccines and immunoinformatics approaches
  • Physical Sciences › Chemistry › Spectroscopy › Advanced Proteomics Techniques and Applications
  • Health Sciences › Medicine › Oncology › Cancer Immunotherapy and Biomarkers
Keywords
immunopeptidomics, CDK4/6 inhibition, absolute quantification, melanoma, interferon-gamma, antigen presentation, cancer immunotherapy
MeSH terms
Antigens, Cell Line, Humans, Immunotherapy, Interferon-gamma, Peptides, Mass Spectrometry, Histocompatibility Antigens Class I, Antigen Presentation, Proteomics, Cyclin-Dependent Kinase 4, Cyclin-Dependent Kinase 6

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