CoChat

Structure of SALL4 zinc finger domain reveals link between AT-rich DNA binding and Okihiro syndrome

Open access

Published version of a preprint

The DOI matches its Crossref record; 53 of the 54 references it cites were found in a registry or an index; 26 of the 54 sources its claims cite were checked for support.

Abstract

Spalt-like 4 (SALL4) maintains vertebrate embryonic stem cell identity and is required for the development of multiple organs, including limbs. Mutations in SALL4 are associated with Okihiro syndrome, and SALL4 is also a known target of thalidomide. SALL4 protein has a distinct preference for AT-rich sequences, recognised by a pair of zinc fingers at the C-terminus. However, unlike many characterised zinc finger proteins, SALL4 shows flexible recognition with many different combinations of AT-rich sequences being targeted. SALL4 interacts with the NuRD corepressor complex which potentially mediates repression of AT-rich genes. We present a crystal structure of SALL4 C-terminal zinc fingers with an AT-rich DNA sequence, which shows that SALL4 uses small hydrophobic and polar side chains to provide flexible recognition in the major groove. Missense mutations reported in patients that lie within the C-terminal zinc fingers reduced overall binding to DNA but not the preference for AT-rich sequences. Furthermore, these mutations altered association of SALL4 with AT-rich genomic sites, providing evidence that these mutations are likely pathogenic.

Overview

CoChat Summary

TL;DR

The paper shows that the SALL4 ZFC4 zinc-finger domain binds AT-rich DNA through a conserved interface of small polar and hydrophobic residues, with an X-ray crystal structure and biochemical tests explaining sequence preference. It also shows that patient missense mutations in ZFC4 reduce DNA binding and disrupt heterochromatin localization, linking impaired AT-rich DNA recognition to Okihiro syndrome.

Key findings

  • The ZFC4–DNA crystal structure was solved at 2.76 Å resolution, with final Rwork/Rfree values of 24.7%/25.4%, and showed ZFC4 contacting bases in the major groove while DNA adopts a compressed minor groove and slightly expanded major groove.
  • ZFC4 recognizes AT-rich sequences using small polar/hydrophobic residues; in binding assays, WT protein had an apparent Kd of 0.76 µM, while I897S and the I897S,V925S double mutant had reduced binding (Kd 4.8 and 5.0 µM), whereas V925S retained binding (Kd 0.91 µM).
  • Patient mutations in ZFC4 reduced DNA binding: R900W had an apparent Kd of 23 µM, and G921D binding was too weak to determine; both mutations also caused diffuse nuclear localization instead of DAPI-bright heterochromatin foci in NIH 3T3 cells.
  • HT-SELEX showed WT and mutant ZFC4 proteins still preferred AT-rich motifs, but enrichment was much higher for WT; the top enriched motifs were exclusively A/T and enrichment tracked better with total A/T content than with consecutive A/T runs.

Methods

  • The study combined structural, biochemical, sequencing-based, and cell-based approaches: X-ray crystallography of mouse SALL4 ZFC4 (residues 870–940) bound to a palindromic AT-rich oligonucleotide, SAXS, EMSA, HT-SELEX, and immunofluorescence in cells.
  • For disease-relevant testing, the authors compared WT SALL4 with patient missense variants H898R, R900W, and G921D in NIH 3T3 fibroblasts, which lack endogenous SALL4 and SALL1.
  • HT-SELEX was run in triplicate, with mutant and WT ZFC4 proteins, and analyzed by k-mer enrichment across cycles 1, 3, and 6; structural comparisons and binding analyses were used to interpret sequence specificity.

Limitations

  • The crystal structure had high anisotropy and limited resolution (2.76 Å), and the authors note they could not fit structured water at the DNA-protein interface.
  • The paper’s binding conclusions for some mutants are constrained by assay limits: for G921D, an EMSA dissociation constant could not be determined.
  • Evident from the text: the work tests only a limited set of mutations and a single ZFC4 DNA motif/oligonucleotide context in the structural assay, so it does not establish all possible sequence or mutation effects.

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

Integrity signals

Registry

Crossref · 10.26508/lsa.202201588

Cites

53 of 53 references verified0 citation errors · 0 not checkable

Statistics

Nothing to recompute.

Figures

Figure 1. SALL4 ZFC4 in complex with DNA. (A) Domain overview of human SALL4 protein. Missense variants from gnomAD population data (blue) are placed relative to the sequence. Above are pathogenic missense mutations (red) and variants of unknown significance (orange) that are absent from the gnomAD database. The NuRD binding region (aa1–12) and Q-rich sequences (aa207–242) are followed by seven zinc finger domains arranged in three clusters (ZFC1, aa382–432; ZFC2, aa566–648; and ZFC4, aa870–920). The variant depletion value for SALL4 (Vp) is given, along with domain-level variant depletion values (VdVp ratios) under the domain labels. A construct of ZFC4 containing zinc fingers 6 (green) and 7 (blue) was used for structural analysis. (B) The palindromic DNA sequence co-crystallised with SALL4. Strand A and B are coloured yellow and white, respectively. (C) An overview of the mouse SALL4 ZFC4–DNA complex, showing a single protein chain interacting with the DNA. Colour schemes match those of schematics in (A, B). (D) An all-atom model of SALL4–ZFC4 complex with DNA fitted into an SAXS envelope. The complex is rotated 90° around the vertical axis relative to (C). Source data are available for this figure.
Figure 1. SALL4 ZFC4 in complex with DNA. (A) Domain overview of human SALL4 protein. Missense variants from gnomAD population data (blue) are placed relative to the sequence. Above are pathogenic missense mutations (red) and variants of unknown significance (orange) that are absent from the gnomAD database. The NuRD binding region (aa1–12) and Q-rich sequences (aa207–242) are followed by seven zinc finger domains arranged in three clusters (ZFC1, aa382–432; ZFC2, aa566–648; and ZFC4, aa870–920). The variant depletion value for SALL4 (Vp) is given, along with domain-level variant depletion values (VdVp ratios) under the domain labels. A construct of ZFC4 containing zinc fingers 6 (green) and 7 (blue) was used for structural analysis. (B) The palindromic DNA sequence co-crystallised with SALL4. Strand A and B are coloured yellow and white, respectively. (C) An overview of the mouse SALL4 ZFC4–DNA complex, showing a single protein chain interacting with the DNA. Colour schemes match those of schematics in (A, B). (D) An all-atom model of SALL4–ZFC4 complex with DNA fitted into an SAXS envelope. The complex is rotated 90° around the vertical axis relative to (C). Source data are available for this figure.CC BY · PubMed Central
Figure S1. Structural and SAXS analysis of SALL4 and DNA. (A) Superposition of all four ZFC4 chains found in the asymmetric unit of the crystal. (B) Superposition of refined AT-rich DNA (white and yellow) compared with ideal B-form (black). (C) SEC-SAXS profiles for SALL4 ZFC4 (teal), dsDNA (grey), and ZFC4–DNA complex (black). (C, D) Real space P(r) functions calculated for SAXS experiments in (C). (E) SAXS curves for samples from (C) showing comparison with curves calculated from all-atom models of SALL4 ZFC4 (teal), DNA (grey), and the ZFC4–DNA complex (black). Guinier analysis for all samples is shown below. (F) Normalised Kratky analysis for SAXS samples. Source data are available for this figure.
Figure S1. Structural and SAXS analysis of SALL4 and DNA. (A) Superposition of all four ZFC4 chains found in the asymmetric unit of the crystal. (B) Superposition of refined AT-rich DNA (white and yellow) compared with ideal B-form (black). (C) SEC-SAXS profiles for SALL4 ZFC4 (teal), dsDNA (grey), and ZFC4–DNA complex (black). (C, D) Real space P(r) functions calculated for SAXS experiments in (C). (E) SAXS curves for samples from (C) showing comparison with curves calculated from all-atom models of SALL4 ZFC4 (teal), DNA (grey), and the ZFC4–DNA complex (black). Guinier analysis for all samples is shown below. (F) Normalised Kratky analysis for SAXS samples. Source data are available for this figure.CC BY · PubMed Central
Figure 2. SALL4 ZFC4 binds DNA with polar and hydrophobic contacts. (A) Alignment of Znf6 and Znf7 showing standard position annotation for zinc finger helix residues along with secondary structure elements (arrows are beta strands, cylinder is an alpha helix) below. (B) Schematic overview of direct interactions between SALL4 ZFC4 and AT-rich DNA. Grey circles represent the methyl groups on thymine bases that point into the major groove. (C) Overview of structure showing positions of zoomed views. Lower panel is related to upper panel by a 90° rotation, matching the views in Fig 1C and D. (D, E, F) Zoomed in views of side chain interactions with AT-rich DNA.
Figure 2. SALL4 ZFC4 binds DNA with polar and hydrophobic contacts. (A) Alignment of Znf6 and Znf7 showing standard position annotation for zinc finger helix residues along with secondary structure elements (arrows are beta strands, cylinder is an alpha helix) below. (B) Schematic overview of direct interactions between SALL4 ZFC4 and AT-rich DNA. Grey circles represent the methyl groups on thymine bases that point into the major groove. (C) Overview of structure showing positions of zoomed views. Lower panel is related to upper panel by a 90° rotation, matching the views in Fig 1C and D. (D, E, F) Zoomed in views of side chain interactions with AT-rich DNA.CC BY · PubMed Central
Figure S2. Features of SALL4 ZFC4 are found in SALL1 and SALL3. (A) 2Fo-Fc map contoured at 1σ around the TGEKP sequence that connects Znf6 to Znf7. R900 of Znf6 hydrogen bonds with the backbone carbonyl group of T918. (B) Refined SALL4 ZFC4 chain L with palindromic DNA, showing the final 2Fo-Fc map contoured at 1σ. The view is of a plane through the A4(yellow)–T9(white) base pair. (C) Multiple protein sequence alignments of SALL4, SALL1, and SALL3 across vertebrates (Homo sapiens [HUMAN], Mus musculus [MOUSE], Gallus gallus [CHICK], Anolis carolinensis [ANOCA], Xenopus tropicalis [XENTR], Danio rerio [DANRE]), with Drosophila melanogaster Salr for reference. The sequences were aligned using MAFFT. Znf6 is highlighted green and Znf7 is highlighted blue. Reference sequence positions for each zinc finger are given above the alignment. DNA-binding residues are boxed in green and blue. Patient missense mutations are boxed in pink and orange and annotated above with human and (mouse) numbering. Secondary structure elements are shown below the sequence. Asterisks indicate residues that have previously been shown to alter SALL4 function in mouse cells. (D) Analysis of gnomAD variants (blue vertical lines) for SALL1 and SALL3 with domain organisation and numbering for domain boundaries below. The Vp values for SALL1 and SALL3 are given below the protein name, whereas VdVp values are given for the calculated domain boundaries. Source data are available for this figure.
Figure S2. Features of SALL4 ZFC4 are found in SALL1 and SALL3. (A) 2Fo-Fc map contoured at 1σ around the TGEKP sequence that connects Znf6 to Znf7. R900 of Znf6 hydrogen bonds with the backbone carbonyl group of T918. (B) Refined SALL4 ZFC4 chain L with palindromic DNA, showing the final 2Fo-Fc map contoured at 1σ. The view is of a plane through the A4(yellow)–T9(white) base pair. (C) Multiple protein sequence alignments of SALL4, SALL1, and SALL3 across vertebrates (Homo sapiens [HUMAN], Mus musculus [MOUSE], Gallus gallus [CHICK], Anolis carolinensis [ANOCA], Xenopus tropicalis [XENTR], Danio rerio [DANRE]), with Drosophila melanogaster Salr for reference. The sequences were aligned using MAFFT. Znf6 is highlighted green and Znf7 is highlighted blue. Reference sequence positions for each zinc finger are given above the alignment. DNA-binding residues are boxed in green and blue. Patient missense mutations are boxed in pink and orange and annotated above with human and (mouse) numbering. Secondary structure elements are shown below the sequence. Asterisks indicate residues that have previously been shown to alter SALL4 function in mouse cells. (D) Analysis of gnomAD variants (blue vertical lines) for SALL1 and SALL3 with domain organisation and numbering for domain boundaries below. The Vp values for SALL1 and SALL3 are given below the protein name, whereas VdVp values are given for the calculated domain boundaries. Source data are available for this figure.CC BY · PubMed Central
Figure 3. Mutation of small hydrophobic residues alters SALL4 binding to DNA. (A) Diagram showing the two residues that were mutated in position 6 of the zinc finger motif. (B) EMSA showing WT SALL4 and altered DNA binding interactions of the mutants. Protein concentrations used in titration points are shown below the gel. (C) Binding curves for SALL4 WT and mutants based on EMSA data. (D) Immunofluorescence of SALL4 WT and mutant proteins in 3T3 cells transfected with expression constructs, with DAPI staining for comparison. Scale bars are 5 µm. (E) Quantification of cells with SALL4 localisation to DAPI foci. The number of cells analysed in each case is given at the top of the chart. Each data point is from an independent transfection experiment. Source data are available for this figure.
Figure 3. Mutation of small hydrophobic residues alters SALL4 binding to DNA. (A) Diagram showing the two residues that were mutated in position 6 of the zinc finger motif. (B) EMSA showing WT SALL4 and altered DNA binding interactions of the mutants. Protein concentrations used in titration points are shown below the gel. (C) Binding curves for SALL4 WT and mutants based on EMSA data. (D) Immunofluorescence of SALL4 WT and mutant proteins in 3T3 cells transfected with expression constructs, with DAPI staining for comparison. Scale bars are 5 µm. (E) Quantification of cells with SALL4 localisation to DAPI foci. The number of cells analysed in each case is given at the top of the chart. Each data point is from an independent transfection experiment. Source data are available for this figure.CC BY · PubMed Central
Figure S3. Purification and characterisation of SALL4 ZFC4 with DNA binding mutations. (A) SDS–PAGE gel (4–20%) with WT and mutant SALL4 ZFC4 (residues 870–940). (B) Superposition of size exclusion chromatograms (from Superdex S75 column), showing elution profiles for V925S (black), I897S (blue), and double mutant (grey) SALL4 ZFC4. Traces at 280 nm have solid lines and 260 nm have dotted lines. Source data are available for this figure.
Figure S3. Purification and characterisation of SALL4 ZFC4 with DNA binding mutations. (A) SDS–PAGE gel (4–20%) with WT and mutant SALL4 ZFC4 (residues 870–940). (B) Superposition of size exclusion chromatograms (from Superdex S75 column), showing elution profiles for V925S (black), I897S (blue), and double mutant (grey) SALL4 ZFC4. Traces at 280 nm have solid lines and 260 nm have dotted lines. Source data are available for this figure.CC BY · PubMed Central
Figure 4. ZFC4 patient missense mutations reduce SALL4 binding to DNA and alter localisation in nuclei. (A) Diagram showing SALL4 ZFC4 WT and mutant constructs used in this study. (B) Model of point mutations H898R and R900W (mouse numbering, pink carbon atoms) superposed on WT structure to show alterations in proteins structure. H898R would disrupt zinc ion binding. R900W likely disrupts the interface between Znf6 and Znf7. (C) Zoomed in view showing the position of the G921D mutation in the major groove. (D) EMSAs showing binding interactions of SALL4 ZFC4 WT and mutant proteins with an AT-rich DNA motif. Protein concentrations used in titration points are shown below the gel. (E) Binding curves for SALL4 WT and mutants based on EMSA data. (F) Immunofluorescence of SALL4 WT and mutant proteins in 3T3 cells transfected with expression constructs, with DAPI staining for comparison. Scale bars are 5 µm. (G) Quantification of cells with SALL4 localisation to DAPI foci. The number of cells analysed in each case is given at the top of the chart. Each data point is from an independent transfection experiment. Source data are available for this figure.
Figure 4. ZFC4 patient missense mutations reduce SALL4 binding to DNA and alter localisation in nuclei. (A) Diagram showing SALL4 ZFC4 WT and mutant constructs used in this study. (B) Model of point mutations H898R and R900W (mouse numbering, pink carbon atoms) superposed on WT structure to show alterations in proteins structure. H898R would disrupt zinc ion binding. R900W likely disrupts the interface between Znf6 and Znf7. (C) Zoomed in view showing the position of the G921D mutation in the major groove. (D) EMSAs showing binding interactions of SALL4 ZFC4 WT and mutant proteins with an AT-rich DNA motif. Protein concentrations used in titration points are shown below the gel. (E) Binding curves for SALL4 WT and mutants based on EMSA data. (F) Immunofluorescence of SALL4 WT and mutant proteins in 3T3 cells transfected with expression constructs, with DAPI staining for comparison. Scale bars are 5 µm. (G) Quantification of cells with SALL4 localisation to DAPI foci. The number of cells analysed in each case is given at the top of the chart. Each data point is from an independent transfection experiment. Source data are available for this figure.CC BY · PubMed Central
Figure S4. Purification and characterisation of SALL4 ZFC4 with patient mutations. (A) SDS–PAGE gel (4–20%) with WT and mutant SALL4 ZFC4 (residues 870–940). (B) Superposition of size exclusion chromatograms (from Superdex S75 column), showing elution profiles for WT (black), G921D (orange), and R900W (pale orange) SALL4 ZFC4. Source data are available for this figure.
Figure S4. Purification and characterisation of SALL4 ZFC4 with patient mutations. (A) SDS–PAGE gel (4–20%) with WT and mutant SALL4 ZFC4 (residues 870–940). (B) Superposition of size exclusion chromatograms (from Superdex S75 column), showing elution profiles for WT (black), G921D (orange), and R900W (pale orange) SALL4 ZFC4. Source data are available for this figure.CC BY · PubMed Central
Figure S5. Characterisation of SALL4, SALL1 and mutant localisation in cells. (A) Plasmid used for expression of WT and mutant SALL4 proteins in 3T3 cells. (B) DAPI staining and immunofluorescence for SALL4 and SALL1 in mouse embryonic stem cells and NIH 3T3 cells. Scale bars are 5 µm. (C) Comparison of SALL4 localisation in cells with different expression levels because of variations in transfection across cell population. Three levels of brightness are shown to enable comparison.
Figure S5. Characterisation of SALL4, SALL1 and mutant localisation in cells. (A) Plasmid used for expression of WT and mutant SALL4 proteins in 3T3 cells. (B) DAPI staining and immunofluorescence for SALL4 and SALL1 in mouse embryonic stem cells and NIH 3T3 cells. Scale bars are 5 µm. (C) Comparison of SALL4 localisation in cells with different expression levels because of variations in transfection across cell population. Three levels of brightness are shown to enable comparison.CC BY · PubMed Central
Figure 5. ZFC4 patient missense mutations do not alter sequence preference. (A) Diagram summarising the HT-SELEX procedure to determine ZFC4 binding specificity. (B) Relative enrichment of 6-mer DNA motifs categorised by total number of A/Ts at cycle 1, 3, and 6 of HT-SELEX with SALL4 ZFC4 WT (blue), R900W (purple), G921D (green), and negative control (grey) samples. Error bars indicate the variability (SD) in three independent replicate experiments. (C) Venn diagram showing the overlap of the top 100 enriched 6-mer DNA motifs at cycle 6 of HT-SELEX with SALL4 ZFC4 WT and mutant proteins. Source data are available for this figure.
Figure 5. ZFC4 patient missense mutations do not alter sequence preference. (A) Diagram summarising the HT-SELEX procedure to determine ZFC4 binding specificity. (B) Relative enrichment of 6-mer DNA motifs categorised by total number of A/Ts at cycle 1, 3, and 6 of HT-SELEX with SALL4 ZFC4 WT (blue), R900W (purple), G921D (green), and negative control (grey) samples. Error bars indicate the variability (SD) in three independent replicate experiments. (C) Venn diagram showing the overlap of the top 100 enriched 6-mer DNA motifs at cycle 6 of HT-SELEX with SALL4 ZFC4 WT and mutant proteins. Source data are available for this figure.CC BY · PubMed Central
Figure S6. Preferential binding sequences of SALL4 from SELEX experiments. (A) Position frequency matrix motif logos of the most enriched 6-mer DNA motifs at cycle 6 of HT-SELEX with SALL4 ZFC4. (B) Relative enrichment (fold change versus cycle 0) of 6-mer DNA motifs categorised by consecutive or total number of A/Ts (see example above the plot) at cycle 6 of HT-SELEX with SALL4 ZFC4 WT. Error bars indicate the variability (SD) in three independent replicate experiments.
Figure S6. Preferential binding sequences of SALL4 from SELEX experiments. (A) Position frequency matrix motif logos of the most enriched 6-mer DNA motifs at cycle 6 of HT-SELEX with SALL4 ZFC4. (B) Relative enrichment (fold change versus cycle 0) of 6-mer DNA motifs categorised by consecutive or total number of A/Ts (see example above the plot) at cycle 6 of HT-SELEX with SALL4 ZFC4 WT. Error bars indicate the variability (SD) in three independent replicate experiments.CC BY · PubMed Central

Verification

Claims the paper makes

18 Valid · 4 Not valid · 18 Not checked · 98 Own results and interpretations

81 more claims · 8 with their cited sources checked

Create a free account to see all 138 claims and to verify the 4 references still open.

See all claims — free account
  • It may state two separate claims as one. 44
  • Probably not a substantial claim. 11

Create a free account to see these claims.

Citations

Cites

Show all 53 references

Research works printed but not linked: 1

  1. Tickle IJ , Flensburg C , Keller P , Paciorek W , Sharff A , Vonrhein C , Bricogne G (2018) STARANISO. Cambridge, United Kingdom: Global Phasing Ltd.

Cited by

See all 14

Metadata

Authors
Watson, J. A., Pantier, R., Jayachandran, U., Chhatbar, K., Alexander-Howden, B., Kruusvee, V., Prendecki, M., Bird, A., & Cook, A. G.
Affiliations
  • Wellcome Centre for Cell Biology (GB)
  • University of Copenhagen (DK)
  • Wellcome Centre for Cell Biology, Max Born Crescent, Edinburgh, UK
Corresponding author
Cook, Atlanta G
Venue
Life Science Alliance
Published
2023-01-12
Volume
6
Issue
3
Pages
e202201588
ISSN
2575-1077 (ISSN-L)
OpenAlex
W4315705236
Cited by
14
References
54
Citations by year
2026: 3 · 2025: 9 · 2024: 1 · 2023: 1
FWCI
1.24
Type
journal article
Language
en
Open access
Locations
  • Life Science Alliance (publishedVersion, cc-by)
  • bioRxiv (Cold Spring Harbor Laboratory) (acceptedVersion, cc-by)
  • PubMed Central (publishedVersion, cc-by)
  • PubMed (publishedVersion)
  • Research at the University of Copenhagen (University of Copenhagen) (publishedVersion, cc-by)
  • HAL (submittedVersion, cc-by)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • Edinburgh Research Explorer (University of Edinburgh) (submittedVersion, cc-by)
Funders
  • Wellcome Trust (200898, 107930, 092076, 203149)
  • EC | ERC | HORIZON EUROPE European Research Council (EC 694295 Gen-Epix)
Topics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Genomics and Chromatin Dynamics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Protein Structure and Dynamics
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Structural Biology › Enzyme Structure and Function
Keywords
SALL4, zinc finger, AT-rich DNA binding, crystal structure, limb development, missense mutations, DNA binding specificity, major groove recognition
MeSH terms
Duane Retraction Syndrome, Humans, Mutation, Transcription Factors, Zinc Fingers

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 2026-10-10, under version 3 of the claim definition (model deepseek-v4.1-flash, prompt extract-v3).

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

Cited sources are checked with a language model’s help (gpt-6-sol).

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

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.