Student or educator? Sign up with your school email and get 6 months of Pro free.
CoChat

The DOI matches its Crossref record; 63 of the 63 references it cites were found in a registry or an index; 18 of the 64 sources its claims cite were checked for support.

Nanoproteomics enables proteoform-resolved analysis of low-abundance proteins in human serum

Open access

Abstract

Top-down mass spectrometry (MS)-based proteomics provides a comprehensive analysis of proteoforms to achieve a proteome-wide understanding of protein functions. However, the MS detection of low-abundance proteins from blood remains an unsolved challenge due to the extraordinary dynamic range of the blood proteome. Here, we develop an integrated nanoproteomics method coupling peptide-functionalized superparamagnetic nanoparticles (NPs) with top-down MS for the enrichment and comprehensive analysis of cardiac troponin I (cTnI), a gold-standard cardiac biomarker, directly from serum. These NPs enable the sensitive enrichment of cTnI (<1 ng/mL) with high specificity and reproducibility, while simultaneously depleting highly abundant proteins such as human serum albumin (>10 10 more abundant than cTnI). We demonstrate that top-down nanoproteomics can provide high-resolution proteoform-resolved molecular fingerprints of diverse cTnI proteoforms to establish proteoform-pathophysiology relationships. This scalable and reproducible antibody-free strategy can generally enable the proteoform-resolved analysis of low-abundance proteins directly from serum to reveal previously unachievable molecular details.

Overview

CoChat Summary

TL;DR

The paper develops an antibody-free top-down nanoproteomics method that uses peptide-functionalized superparamagnetic nanoparticles to enrich cardiac troponin I (cTnI) directly from human serum and analyze its proteoforms by top-down mass spectrometry. The approach enriches cTnI while depleting highly abundant human serum albumin and detects cTnI proteoforms at clinically relevant low ng/mL levels.

Key findings

  • The NP-Pep uses a 13-mer cTnI-binding peptide (HWQIAYNEHQWQC) with a reported Kd of 270 pM, targeting cTnI residues 114–144; peptide surface density was inferred as ~0.034 µmol peptide/mg NP, and NP-Pep zeta potential in 0.1× PBS was ~−38 mV.
  • From human heart sarcomeric extracts, NP-Pep enriched cTnI over threefold (p < 0.0001) relative to the loading mixture and preserved endogenous cTnI proteoform distributions, including ppcTnI[1-207], cTnI, pcTnI, and ppcTnI.
  • From human serum, NP-Pep enabled detection of cTnI proteoforms with an LOD of 0.75 ng/mL, while top-down MS alone detected purified cTnI with an LOD of 0.06 ng/mL; NP-Pep showed a 115-fold cTnI enrichment factor and ~51% serum cTnI recovery, compared with ~17% for Agarose-mAb.
  • NP-Pep nearly completely depleted HSA, which is >10 orders of magnitude more abundant than circulating cTnI, and revealed proteoform fingerprints: donor hearts were primarily phosphorylated cTnI, dilated cardiomyopathy hearts showed decreased phosphorylated cTnI, and postmortem samples showed C-terminal truncated cTnI proteoforms.

Methods

  • The authors synthesized BAPTES-silanized Fe3O4 nanoparticles and conjugated the cTnI-binding peptide via cysteine-thiol chemistry, characterizing NPs by TEM, FTIR, TGA, and zeta potential.
  • cTnI enrichment was tested from human heart sarcomeric extracts and from human male AB serum spiked with cTnI from donor, dilated cardiomyopathy, and postmortem heart samples; NP-Pep was compared with agarose-peptide and anti-cTnI monoclonal antibody (M46) platforms.
  • Enriched fractions were analyzed by SDS-PAGE, ELISA, and top-down LC/MS using reversed-phase liquid chromatography coupled to a maXis II Q-TOF mass spectrometer, with deconvolution and MS/MS for proteoform identification.

Limitations

  • The paper states that some nonspecific proteins were still retained from serum enrichment, that cTnI recovery could be improved by automating the NP-Pep workflow to reduce sample handling and protein loss, and that further instrumentation improvements are needed to reach the diagnostic cutoff used by contemporary cTnI ELISA (≤0.04 ng/mL).
  • The paper notes that the allene carboxamide chemistry relies on a Cys nonessential to the peptide binding sequence, with alternative bioorthogonal coupling approaches needed if such a Cys is unavailable.
  • Evident from the text: serum experiments used spike-in cTnI from a limited set of human heart samples rather than endogenous circulating cTnI from a large patient cohort, so clinical validation remains future work.

Model-written summary. Check it against the paper before relying on it.

Integrity signals

Registry

Crossref · 10.1038/s41467-020-17643-1

Cites

63 of 63 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Fig. 1 Design and characterization of surface-functionalized nanoparticles (NPs) for capturing cTnI. a Silanization of Fe3O4 NPs using an allene carboxamide-based organosilane monomer (BAPTES) for cysteine-thiol-specific bioconjugation. b Illustration of the rationally designed NPs that are surface functionalized with a 13-mer peptide that has a high affinity for cTnI (NP-Pep) for cTnI enrichment. The 13-mer peptide possesses a C-terminal cysteine that selectively reacts with the allene carboxamide moiety on the silanized NPs. Representative TEM images of surface-functionalized NPs: Fe3O4-OA NPs (c) (inset shows the selected area electron diffraction pattern), Fe3O4-BAPTES NPs (d), and Fe3O4-Peptide NPs (e). Representative FTIR spectra (f) and TGA analysis (g) of various NPs: Fe3O4-OA NPs (black), Fe3O4-BAPTES NPs (red), and Fe3O4-peptide NPs (blue). FTIR spectra are offset along the y-axis for clarity, with arrows denoting the characteristic allene (C = C = C) vibrational modes (1971 and 1947 cm−1) corresponding to the silanized NPs. TGA analysis reveals increased organic content on NPs after peptide coupling. h Photographs of functionalized NPs in a biphasic mixture of dichloromethane (CH2Cl2) and water (H2O), comparing the solvent compatibility of the NP-BAPTES and the NP-Pep. The NP-BAPTES are dispersible in dichloromethane but the NP-Pep are stable and dispersible in water. The displayed NP-Pep and NP-BAPTES originated from the same synthetic batch. NP-BAPTES is referred to as NP-Ctrl in the subsequent figures and text. Data are representative of n = 3 independent experiments. Source data are provided as a Source Data file.
Fig. 1 Design and characterization of surface-functionalized nanoparticles (NPs) for capturing cTnI. a Silanization of Fe3O4 NPs using an allene carboxamide-based organosilane monomer (BAPTES) for cysteine-thiol-specific bioconjugation. b Illustration of the rationally designed NPs that are surface functionalized with a 13-mer peptide that has a high affinity for cTnI (NP-Pep) for cTnI enrichment. The 13-mer peptide possesses a C-terminal cysteine that selectively reacts with the allene carboxamide moiety on the silanized NPs. Representative TEM images of surface-functionalized NPs: Fe3O4-OA NPs (c) (inset shows the selected area electron diffraction pattern), Fe3O4-BAPTES NPs (d), and Fe3O4-Peptide NPs (e). Representative FTIR spectra (f) and TGA analysis (g) of various NPs: Fe3O4-OA NPs (black), Fe3O4-BAPTES NPs (red), and Fe3O4-peptide NPs (blue). FTIR spectra are offset along the y-axis for clarity, with arrows denoting the characteristic allene (C = C = C) vibrational modes (1971 and 1947 cm−1) corresponding to the silanized NPs. TGA analysis reveals increased organic content on NPs after peptide coupling. h Photographs of functionalized NPs in a biphasic mixture of dichloromethane (CH2Cl2) and water (H2O), comparing the solvent compatibility of the NP-BAPTES and the NP-Pep. The NP-BAPTES are dispersible in dichloromethane but the NP-Pep are stable and dispersible in water. The displayed NP-Pep and NP-BAPTES originated from the same synthetic batch. NP-BAPTES is referred to as NP-Ctrl in the subsequent figures and text. Data are representative of n = 3 independent experiments. Source data are provided as a Source Data file.CC BY · PubMed Central
Fig. 2 Evaluation of enrichment performance and reproducibility of NP-Pep for capturing cTnI. a Schematic illustration of the nanoproteomics strategy for cTnI enrichment and top-down MS analysis of cTnI proteoforms. Heart tissue extract (loading mixture, L) is first incubated with the NP-Pep. Following magnetic isolation, the nonspecific proteins are removed as flow through (F). The NPs are washed and the NP-bound proteins of interest are eluted (E) and analyzed by top-down LC/MS. b SDS-PAGE demonstrates the high reproducibility obtained from three different NP-Pep syntheses (inter-batch). The lower panels highlight the specific enrichment of cTnI and cTnT in E as compared to L and F. Evaluation of total cTnI proteoform recovery (c), and relative abundance of cTnI (d), before and after NP-Pep enrichment. The same L, F, and E shown in the SDS-PAGE from (b) were equally loaded (500 ng) for LC/MS analysis. The deconvoluted top-down mass spectra corresponding to cTnI proteoforms were used to calculate the relative abundance of each cTnI proteoform when normalizing for total protein amount injected. Proteoform abundance data (c) are representative of n = 6 independent experiments with error bars indicating standard error of the mean. cTnI relative abundance data (d) are representative of n = 3 independent experiments with error bars indicating standard error of the mean. e MS-based evaluation of cTnI enrichment using three different synthetic batches of NP-Pep showing the reproducible enrichment performance. f–h Enrichment from three different human heart samples using NP-Pep in comparison with agarose-mAb. Upper panels feature SDS-PAGE strips that visualize cTnI enrichment performance of NP-Pep and agarose-mAb. Equal protein amount (500 ng) of the L, F, and E was loaded on the gel in (b, f–h). Roman numerals correspond to N-terminally acetylated cTnI proteoforms following Met exclusion: (i) ppcTnI[1-207]; (ii) cTnI; (iii) pcTnI; (iv) ppcTnI; (v) cTnI[1-205]; (vi) pcTnI[1-205]; (vii) cTnI[1-206]; (viii) pcTnI[1-206]. M. protein marker, Std. endogenous cTnI protein standard, p phosphorylation, pp bisphosphorylation. A summary of the enriched proteoforms and their respective mass measurements by top-down MS are listed in Supplementary Tables 5 and 6. Source data are provided as a Source Data file.
Fig. 2 Evaluation of enrichment performance and reproducibility of NP-Pep for capturing cTnI. a Schematic illustration of the nanoproteomics strategy for cTnI enrichment and top-down MS analysis of cTnI proteoforms. Heart tissue extract (loading mixture, L) is first incubated with the NP-Pep. Following magnetic isolation, the nonspecific proteins are removed as flow through (F). The NPs are washed and the NP-bound proteins of interest are eluted (E) and analyzed by top-down LC/MS. b SDS-PAGE demonstrates the high reproducibility obtained from three different NP-Pep syntheses (inter-batch). The lower panels highlight the specific enrichment of cTnI and cTnT in E as compared to L and F. Evaluation of total cTnI proteoform recovery (c), and relative abundance of cTnI (d), before and after NP-Pep enrichment. The same L, F, and E shown in the SDS-PAGE from (b) were equally loaded (500 ng) for LC/MS analysis. The deconvoluted top-down mass spectra corresponding to cTnI proteoforms were used to calculate the relative abundance of each cTnI proteoform when normalizing for total protein amount injected. Proteoform abundance data (c) are representative of n = 6 independent experiments with error bars indicating standard error of the mean. cTnI relative abundance data (d) are representative of n = 3 independent experiments with error bars indicating standard error of the mean. e MS-based evaluation of cTnI enrichment using three different synthetic batches of NP-Pep showing the reproducible enrichment performance. f–h Enrichment from three different human heart samples using NP-Pep in comparison with agarose-mAb. Upper panels feature SDS-PAGE strips that visualize cTnI enrichment performance of NP-Pep and agarose-mAb. Equal protein amount (500 ng) of the L, F, and E was loaded on the gel in (b, f–h). Roman numerals correspond to N-terminally acetylated cTnI proteoforms following Met exclusion: (i) ppcTnI[1-207]; (ii) cTnI; (iii) pcTnI; (iv) ppcTnI; (v) cTnI[1-205]; (vi) pcTnI[1-205]; (vii) cTnI[1-206]; (viii) pcTnI[1-206]. M. protein marker, Std. endogenous cTnI protein standard, p phosphorylation, pp bisphosphorylation. A summary of the enriched proteoforms and their respective mass measurements by top-down MS are listed in Supplementary Tables 5 and 6. Source data are provided as a Source Data file.CC BY · PubMed Central
Fig. 3 Direct enrichment of cTnI from human serum by NP-Pep with high specificity and sensitivity. Human serum with spike-in cTnI was used for the evaluation of cTnI enrichment performance using NP-Pep and the conventional agarose-mAb. a SDS-PAGE visualizing the effectiveness of cTnI enrichment from human serum using different affinity platforms. The cTnI (306 ng) was obtained from a sarcomeric protein extract from a healthy donor heart and spiked into human serum (10 mg). Loading mixture (L), flow through (F), and elution mixture after enrichment (E) were equally loaded on the gel (500 ng) using NP-Ctrl, NP-Pep, Agarose-Ctrl, Agarose-Pep, and Agarose-mAb. The NP-Pep enabled nearly complete depletion of human serum albumin (HSA), allowing for more specific enrichment of cTnI from human serum. b Normalized extracted ion chromatograms (EICs) of cTnI (black) and HSA (red) corresponding to the NP-Pep, Agarose-Pep, and Agarose-mAb shown in (a). Normalized deconvoluted mass spectra corresponding to enriched cTnI (c) and depleted HSA (d), illustrating the abundance of cTnI and HSA before and after enrichment using NP-Pep, Agarose-Pep, and Agarose-mAb corresponding to (a, b). e Normalized EICs illustrating the MS response for cTnI (3.1–0.006 ng/ml) obtained from a human heart tissue extract (n = 2 independent experiments) using a Bruker maXis II ETD. f Evaluation of the sensitivity performance of the nanoproteomics assay. MS response against cTnI concentration is shown for various NP-Pep enrichment fractions from serum mixtures spiked with cTnI (22.53–0.50 ng/mL). The deconvoluted peak intensities of ppcTnI were used for the analyses. Data are representative of n = 3 independent experiments with error bars indicating standard error of the mean. g Representative deconvoluted mass spectra illustrating the MS response for ppcTnI corresponding to the plot in (f). M. protein marker, Std. endogenous cTnI protein standard, p phosphorylation, pp bisphosphorylation, HSA human serum albumin, C-HSA cysteinylated human serum albumin, C-C-HSA doubly cysteinylated human serum albumin, NP-Ctrl unfunctionalized NP, NP-Pep high affinity peptide-functionalized NP, Agarose-Ctrl unfunctionalized agarose, Agarose-Pep high affinity peptide-functionalized agarose, Agarose-mAb antibody (mAb M46) functionalized with agarose. Source data are provided as a Source Data file.
Fig. 3 Direct enrichment of cTnI from human serum by NP-Pep with high specificity and sensitivity. Human serum with spike-in cTnI was used for the evaluation of cTnI enrichment performance using NP-Pep and the conventional agarose-mAb. a SDS-PAGE visualizing the effectiveness of cTnI enrichment from human serum using different affinity platforms. The cTnI (306 ng) was obtained from a sarcomeric protein extract from a healthy donor heart and spiked into human serum (10 mg). Loading mixture (L), flow through (F), and elution mixture after enrichment (E) were equally loaded on the gel (500 ng) using NP-Ctrl, NP-Pep, Agarose-Ctrl, Agarose-Pep, and Agarose-mAb. The NP-Pep enabled nearly complete depletion of human serum albumin (HSA), allowing for more specific enrichment of cTnI from human serum. b Normalized extracted ion chromatograms (EICs) of cTnI (black) and HSA (red) corresponding to the NP-Pep, Agarose-Pep, and Agarose-mAb shown in (a). Normalized deconvoluted mass spectra corresponding to enriched cTnI (c) and depleted HSA (d), illustrating the abundance of cTnI and HSA before and after enrichment using NP-Pep, Agarose-Pep, and Agarose-mAb corresponding to (a, b). e Normalized EICs illustrating the MS response for cTnI (3.1–0.006 ng/ml) obtained from a human heart tissue extract (n = 2 independent experiments) using a Bruker maXis II ETD. f Evaluation of the sensitivity performance of the nanoproteomics assay. MS response against cTnI concentration is shown for various NP-Pep enrichment fractions from serum mixtures spiked with cTnI (22.53–0.50 ng/mL). The deconvoluted peak intensities of ppcTnI were used for the analyses. Data are representative of n = 3 independent experiments with error bars indicating standard error of the mean. g Representative deconvoluted mass spectra illustrating the MS response for ppcTnI corresponding to the plot in (f). M. protein marker, Std. endogenous cTnI protein standard, p phosphorylation, pp bisphosphorylation, HSA human serum albumin, C-HSA cysteinylated human serum albumin, C-C-HSA doubly cysteinylated human serum albumin, NP-Ctrl unfunctionalized NP, NP-Pep high affinity peptide-functionalized NP, Agarose-Ctrl unfunctionalized agarose, Agarose-Pep high affinity peptide-functionalized agarose, Agarose-mAb antibody (mAb M46) functionalized with agarose. Source data are provided as a Source Data file.CC BY · PubMed Central
Fig. 4 Nanoproteomics enables comprehensive analysis of cTnI proteoforms from human serum. a Nanoproteomics assay utilizing NP-Pep for specific enrichment of cTnI from serum and subsequent top-down MS analysis of cTnI proteoforms. cTnI is first spiked into human serum to prepare the loading mixture (L). The NPs are then incubated with the serum loading mixture, the cTnI-bound NPs are magnetically isolated, the unwanted and nonspecific proteins are removed as flow through (F). The captured cTnI is then eluted and the final elution fraction after enrichment is analyzed by top-down LC/MS. b Deconvoluted MS corresponding to cTnI proteoforms enriched from human serum. The cTnI (~10–20 ng/mL) spiked in the human serum (10 mg) were extracted from various human hearts: (i) and (ii), donor hearts; (iii) and (iv), diseased hearts with dilated cardiomyopathy; (v) and (vi), postmortem hearts. cTnI proteoforms were identified using accurate intact mass measurement, using the most abundant mass based on the amino acid sequence of entry name TNNI3_human from the UniProtKB sequence database. p phosphorylation, pp bisphosphorylation. Data are representative of n = 3 independent experiments. A summary of the enriched proteoforms from serum and their respective mass measurements by top-down MS are listed in Supplementary Table 8. Source data are provided as a Source Data file.
Fig. 4 Nanoproteomics enables comprehensive analysis of cTnI proteoforms from human serum. a Nanoproteomics assay utilizing NP-Pep for specific enrichment of cTnI from serum and subsequent top-down MS analysis of cTnI proteoforms. cTnI is first spiked into human serum to prepare the loading mixture (L). The NPs are then incubated with the serum loading mixture, the cTnI-bound NPs are magnetically isolated, the unwanted and nonspecific proteins are removed as flow through (F). The captured cTnI is then eluted and the final elution fraction after enrichment is analyzed by top-down LC/MS. b Deconvoluted MS corresponding to cTnI proteoforms enriched from human serum. The cTnI (~10–20 ng/mL) spiked in the human serum (10 mg) were extracted from various human hearts: (i) and (ii), donor hearts; (iii) and (iv), diseased hearts with dilated cardiomyopathy; (v) and (vi), postmortem hearts. cTnI proteoforms were identified using accurate intact mass measurement, using the most abundant mass based on the amino acid sequence of entry name TNNI3_human from the UniProtKB sequence database. p phosphorylation, pp bisphosphorylation. Data are representative of n = 3 independent experiments. A summary of the enriched proteoforms from serum and their respective mass measurements by top-down MS are listed in Supplementary Table 8. Source data are provided as a Source Data file.CC BY · PubMed Central

Verification

Claims the paper makes

7 Valid · 0 Not valid · 16 Not checked · 50 Own results and interpretations

31 more claims · 1 with their cited sources checked

Create a free account to see all 73 claims and to verify the 46 references still open.

See all claims — free account
  • Probably not a substantial claim. 11
  • It may state two separate claims as one. 26
  • The quote only partly states this claim. 1

Create a free account to see these claims.

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.

Each contradiction between two passages listed here was confirmed by an independent check.

2 of 2

Citations

Cites

Show all 63 references

Cited by

See all 76

Metadata

Authors
Tiambeng, T. N., Roberts, D. S., Brown, K. A., Zhu, Y., Chen, B., Wu, Z., Mitchell, S. D., Guardado-Alvarez, T. M., Jin, S., & Ge, Y.
Affiliations
University of Wisconsin–Madison (US)
Corresponding author
  • Jin, Song
  • Ge, Ying
Venue
Nature Communications
Published
2020-08-06
Volume
11
Issue
1
Pages
3903-3903
ISSN
2041-1723 (ISSN-L)
OpenAlex
W3047360653
Cited by
76
References
63
Citations by year
2026: 16 · 2025: 6 · 2024: 10 · 2023: 19 · 2022: 10 · 2021: 14 · 2020: 1
FWCI
4.20
Type
journal article
Language
en
Open access
Licence
Locations
  • Nature Communications (publishedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • RePEc: Research Papers in Economics (submittedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
Funders
  • U.S. Department of Health & Human Services | NIH | National Heart, Lung, and Blood Institute (HL007936-19)
  • U.S. Department of Health & Human Services | NIH | NIH Office of the Director (OD018475)
  • U.S. Department of Health & Human Services | NIH | National Institute of General Medical Sciences (GM117058, GM125085)
Topics
  • Physical Sciences › Chemistry › Spectroscopy › Advanced Proteomics Techniques and Applications
  • Physical Sciences › Chemistry › Spectroscopy › Mass Spectrometry Techniques and Applications
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Advanced Biosensing Techniques and Applications
Keywords
top-down mass spectrometry, low-abundance proteins, human serum, cardiac troponin I, cardiac biomarkers, superparamagnetic nanoparticles, peptide-functionalized nanoparticles, protein enrichment, proteoforms, cardiac pathophysiology
MeSH terms
Serum Albumin, Human, Blood Chemical Analysis, Blood Proteins, Humans, Protein Processing, Post-Translational, Mass Spectrometry, Reproducibility of Results, Biomarkers, Troponin I, Proteome, Nanotechnology, Proteomics, Magnetite Nanoparticles

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.

The claim verdicts this report counts are experimental: their accuracy is not yet measured.

A language model drew the claims from the paper on 2026-10-05, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

A second model labelled each claim on 2026-10-05 (model jev-1.13-20260917, definition revision gold-guideline-v0.2).

Cited sources are checked with a language model’s help (gpt-6-sol and qwen3.8-omni-flash).

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

How to cite

Cite CoChat for the checks on this page, and the paper for its own findings.

This page
CoChat. (2026). Citation check: Nanoproteomics enables proteoform-resolved analysis of low-abundance proteins in human serum. https://app.cochat.ai/papers/10.1038/s41467-020-17643-1
The paper
Tiambeng, T. N., Roberts, D. S., Brown, K. A., Zhu, Y., Chen, B., Wu, Z., Mitchell, S. D., Guardado-Alvarez, T. M., Jin, S., & Ge, Y. (2020). Nanoproteomics enables proteoform-resolved analysis of low-abundance proteins in human serum. Nature Communications. https://doi.org/10.1038/s41467-020-17643-1

Data sources

Bibliographic data: Semantic Scholar, OpenAlex, Crossref.

Contains information from the Semantic Scholar Academic Graph Datasets, made available under the ODC Attribution License (ODC-BY 1.0).

Semantic Scholar’s data is described in: Kinney, R., et al. (2023). The Semantic Scholar Open Data Platform. arXiv:2301.10140.