CoChat

Diagnostic utility of DNA methylation analysis in genetically unsolved pediatric epilepsies and CHD2 episignature refinement

Open access

Published version of a preprint

The DOI matches its Crossref record; 96 of the 96 references it cites were found in a registry or an index.

Abstract

Sequence-based genetic testing identifies causative variants in ~ 50% of individuals with developmental and epileptic encephalopathies (DEEs). Aberrant changes in DNA methylation are implicated in various neurodevelopmental disorders but remain unstudied in DEEs. We interrogate the diagnostic utility of genome-wide DNA methylation array analysis on peripheral blood samples from 582 individuals with genetically unsolved DEEs. We identify rare differentially methylated regions (DMRs) and explanatory episignatures to uncover causative and candidate genetic etiologies in 12 individuals. Using long-read sequencing, we identify DNA variants underlying rare DMRs, including one balanced translocation, three CG-rich repeat expansions, and four copy number variants. We also identify pathogenic variants associated with episignatures. Finally, we refine the CHD2 episignature using an 850 K methylation array and bisulfite sequencing to investigate potential insights into CHD2 pathophysiology. Our study demonstrates the diagnostic yield of genome-wide DNA methylation analysis to identify causal and candidate variants as 2% (12/582) for unsolved DEE cases.

Overview

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

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

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

Integrity signals

Registry

Crossref · 10.1038/s41467-024-50159-6

Cites

96 of 96 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig. 1 Description of the DNA methylation analysis and features of the study cohort. A Flowchart displaying filtering of samples after quality control (QC) and normalization, analysis pipeline for detecting rare DMRs and episignatures, and narrowing down DNA methylation candidates for this study. B Breakdown of the cohort after QC and normalization (n = 1194) containing individuals with genetically unsolved DEEs (uDEEs, n = 582), unaffected controls (n = 461), analytical controls (n = 143), and affected family members (n = 8). C Number of males (n = 638) versus females (n = 556). D Sample type as blood (n = 1189) versus other tissue types (n = 5). Figure 1A was created with BioRender.com and released under a Creative Commons Attribution-Non-Commercial-Noderivs 4.0 International License (CC-BY-NC-ND).
Fig. 1 Description of the DNA methylation analysis and features of the study cohort. A Flowchart displaying filtering of samples after quality control (QC) and normalization, analysis pipeline for detecting rare DMRs and episignatures, and narrowing down DNA methylation candidates for this study. B Breakdown of the cohort after QC and normalization (n = 1194) containing individuals with genetically unsolved DEEs (uDEEs, n = 582), unaffected controls (n = 461), analytical controls (n = 143), and affected family members (n = 8). C Number of males (n = 638) versus females (n = 556). D Sample type as blood (n = 1189) versus other tissue types (n = 5). Figure 1A was created with BioRender.com and released under a Creative Commons Attribution-Non-Commercial-Noderivs 4.0 International License (CC-BY-NC-ND).CC BY · PubMed Central
Fig. 2 Rare outlier DMR analysis identifies chromosomal hypermethylation caused by X;13 translocation. A Graphical representation of chr13 rare hypermethylation events in a proband with uDEE. The upper portion of the track displays the genes on chr13 for which hypermethylation events were called. The two grey panels (upper=line, lower=dot) depict β-values for the average of the proband’s array replicates (red) and the average of the parents’ array data (black) for a representative probe within each DMR (n = 26). Subtle hypermethylation hovering around ~ 25% can be seen for the proband compared to the parents. The lower track shows chromosomal locations of the X;13 translocation. B Pedigree showing that chr13 hypermethylation events and the X;13 translocation occurred de novo. C IGV view of ONT LRS data for chrX (left) and chr13 (right). Some, but not all, reads spanning the translocation are colored to show that they span the breakpoint. Karyotype plots, as shown in Fig. 2A and others throughout the manuscript, were created using the R package karyoploteR98.
Fig. 2 Rare outlier DMR analysis identifies chromosomal hypermethylation caused by X;13 translocation. A Graphical representation of chr13 rare hypermethylation events in a proband with uDEE. The upper portion of the track displays the genes on chr13 for which hypermethylation events were called. The two grey panels (upper=line, lower=dot) depict β-values for the average of the proband’s array replicates (red) and the average of the parents’ array data (black) for a representative probe within each DMR (n = 26). Subtle hypermethylation hovering around ~ 25% can be seen for the proband compared to the parents. The lower track shows chromosomal locations of the X;13 translocation. B Pedigree showing that chr13 hypermethylation events and the X;13 translocation occurred de novo. C IGV view of ONT LRS data for chrX (left) and chr13 (right). Some, but not all, reads spanning the translocation are colored to show that they span the breakpoint. Karyotype plots, as shown in Fig. 2A and others throughout the manuscript, were created using the R package karyoploteR98.CC BY · PubMed Central
Fig. 3 Rare outlier DMR analysis identifies tandem repeat expansions. A DMR plot depicting outlier hypermethylation of the CSNK1E 5’UTR and intron 1 in two probands with uDEEs (three additional replicate samples across both for a total of five samples), one mother, and two unaffected controls (total n = 8 red lines) detected through epivariation analysis. B The upper panel shows expression values from RNA-seq of human-derived fibroblasts for individuals with CSNK1E hypermethylation (n = 3 biological replicates) compared to individuals with control methylation levels (n = 3 biological replicates). Significance between groups was determined by a two-tailed paired t-test (p = 0.029 for gene counts and p = 0.0169 for transcripts per million or TPM). *P < 0.05. A representative predicted expression plot from drop-out analysis using the OUTRIDER algorithm is shown at the lower portion of the panel. See Supplementary Fig. 9 for the individual OUTRIDER plots for each family and significance information. C Unphased IGV view of LRS data showing CpG sites that are methylated (red) and unmethylated (blue). The CGG repeat expansion seen in the proband was inherited from the mother (Family 1) and is shown as purple squares denoting insertions in the reads (black arrows); not all reads that are methylated show the insertion as they terminated within the inserted sequence and are clipped by the alignment process.
Fig. 3 Rare outlier DMR analysis identifies tandem repeat expansions. A DMR plot depicting outlier hypermethylation of the CSNK1E 5’UTR and intron 1 in two probands with uDEEs (three additional replicate samples across both for a total of five samples), one mother, and two unaffected controls (total n = 8 red lines) detected through epivariation analysis. B The upper panel shows expression values from RNA-seq of human-derived fibroblasts for individuals with CSNK1E hypermethylation (n = 3 biological replicates) compared to individuals with control methylation levels (n = 3 biological replicates). Significance between groups was determined by a two-tailed paired t-test (p = 0.029 for gene counts and p = 0.0169 for transcripts per million or TPM). *P < 0.05. A representative predicted expression plot from drop-out analysis using the OUTRIDER algorithm is shown at the lower portion of the panel. See Supplementary Fig. 9 for the individual OUTRIDER plots for each family and significance information. C Unphased IGV view of LRS data showing CpG sites that are methylated (red) and unmethylated (blue). The CGG repeat expansion seen in the proband was inherited from the mother (Family 1) and is shown as purple squares denoting insertions in the reads (black arrows); not all reads that are methylated show the insertion as they terminated within the inserted sequence and are clipped by the alignment process.CC BY · PubMed Central
Fig. 4 Rare outlier DMR analysis identifies copy number deletion in a family with GEFs + . A DMR plot depicting outlier hypermethylation of the STX1B promoter and TSS in a proband with uDEE detected through epivariation analysis. B Pedigree for the immediate family members indicating that the inheritance of the hypermethylation (red) and copy number deletion (detected through genome sequencing of the proband shown in Fig. 3D) resides on the maternal side. The arrow points to the proband. C Sanger sequencing validation of the copy number deletion breakpoints in the proband, affected sister, mother, and maternal grandmother. Inserted “CACC” sequence is present between the mapped breakpoints. The father and unaffected brother had no PCR product at the same reaction conditions, indicating they did not harbor the deletion. Primer pairs with one partner located within the deletion on each side of the breakpoint were used to amplify the wild-type allele as a control (Supplementary Fig. 13). D IGV view of genome sequencing data showing a 1784 bp heterozygous deletion encompassing part of intron 1, exon 1, the TSS, and the promoter of STX1B. The reads are colored by insertion size and pair orientation and viewed as pairs. The red pairs, which span the breakpoints of the deletion, indicate that the insertion size is greater than expected. The “CACC” insertion sequence is present in the soft clipped bases (not shown).
Fig. 4 Rare outlier DMR analysis identifies copy number deletion in a family with GEFs + . A DMR plot depicting outlier hypermethylation of the STX1B promoter and TSS in a proband with uDEE detected through epivariation analysis. B Pedigree for the immediate family members indicating that the inheritance of the hypermethylation (red) and copy number deletion (detected through genome sequencing of the proband shown in Fig. 3D) resides on the maternal side. The arrow points to the proband. C Sanger sequencing validation of the copy number deletion breakpoints in the proband, affected sister, mother, and maternal grandmother. Inserted “CACC” sequence is present between the mapped breakpoints. The father and unaffected brother had no PCR product at the same reaction conditions, indicating they did not harbor the deletion. Primer pairs with one partner located within the deletion on each side of the breakpoint were used to amplify the wild-type allele as a control (Supplementary Fig. 13). D IGV view of genome sequencing data showing a 1784 bp heterozygous deletion encompassing part of intron 1, exon 1, the TSS, and the promoter of STX1B. The reads are colored by insertion size and pair orientation and viewed as pairs. The red pairs, which span the breakpoints of the deletion, indicate that the insertion size is greater than expected. The “CACC” insertion sequence is present in the soft clipped bases (not shown).CC BY · PubMed Central
Fig. 5 Summary methylation variant pathogenicity (MVP) score for all individuals positive for EpisignTM v4 episignature analysis. A Methylation Variant Pathogenicity (MVP) score (between 0 and 1) was generated to represent the confidence of prediction for the specific episignature on the EpiSignTM v4 clinical classifier that the SVM was trained to detect. Each colored circle represents a different individual and its associated MVP score for each of the episignatures on the EpiSignTM v4 clinical classifier. Final classification for a specific EpiSignTM disorder includes a combination of MVP score, hierarchical clustering, and multidimensional scaling (MDS) review.
Fig. 5 Summary methylation variant pathogenicity (MVP) score for all individuals positive for EpisignTM v4 episignature analysis. A Methylation Variant Pathogenicity (MVP) score (between 0 and 1) was generated to represent the confidence of prediction for the specific episignature on the EpiSignTM v4 clinical classifier that the SVM was trained to detect. Each colored circle represents a different individual and its associated MVP score for each of the episignatures on the EpiSignTM v4 clinical classifier. Final classification for a specific EpiSignTM disorder includes a combination of MVP score, hierarchical clustering, and multidimensional scaling (MDS) review.CC BY · PubMed Central
Fig. 6 Insights from the CHD2 episignature. A Multidimensional scaling (MDS) plot showing clustering of individuals with pathogenic CHD2 variants (red, upper) for the previously described CHD2 450 K (n = 9) episignature with shared 450 K and 850 K array probes clusters away from the controls (n = 54, blue). The refined CHD2 850 K (n = 29) episignature (red, lower) clusters away from unaffected controls (n = 58, blue). B Circos plot representing shared probes between episignatures. Differentially methylated probes (DMPs) shared between the CHD2 850 K cohort (bold red), CHD2 450 K cohort (red), and 55 other episignatures on EpiSign with functional correlation analysis previously published31. The thickness of the connecting lines corresponds to the number of probes shared between the cohorts. C Tree and leaf visualization of Euclidean clustering of episignatures. Tree and leaf visualization for all 57 cohorts using the top 500 DMPs for each group (for cohorts with less than 500 DMPs, all DMPS were used). Cohort samples were aggregated using the median value of each probe within a group. A leaf node represents a cohort, with node sizes illustrating relative scales of the number of selected DMPs for the respective cohort, and node colors are indicative of the global mean methylation difference, a gradient of hypomethylation (blue) or hypermethylation (red). D Circular karyotype plot showing overlap of CHD2 450 K episignature probes (inner circle, n = 200), with CHD2 850 K episignature probes (middle circle, n = 200), and WGBS DMRs (n = 4 CHD2 vs. n = 6 unaffected controls) derived with at least a 15% methylation difference for the condensed visual representation (outer circle, n = 411). Each line depicts a probe or DMR where red denotes hypermethylation and blue denotes hypomethylation. The purple tracks depict coverage of the 450 K array probes (inner), 850 K EPIC array probes (middle), and WGBS reads (outer). Refer to Supplementary Fig. 33 for linear karyotype DMR plots for chr1-22.
Fig. 6 Insights from the CHD2 episignature. A Multidimensional scaling (MDS) plot showing clustering of individuals with pathogenic CHD2 variants (red, upper) for the previously described CHD2 450 K (n = 9) episignature with shared 450 K and 850 K array probes clusters away from the controls (n = 54, blue). The refined CHD2 850 K (n = 29) episignature (red, lower) clusters away from unaffected controls (n = 58, blue). B Circos plot representing shared probes between episignatures. Differentially methylated probes (DMPs) shared between the CHD2 850 K cohort (bold red), CHD2 450 K cohort (red), and 55 other episignatures on EpiSign with functional correlation analysis previously published31. The thickness of the connecting lines corresponds to the number of probes shared between the cohorts. C Tree and leaf visualization of Euclidean clustering of episignatures. Tree and leaf visualization for all 57 cohorts using the top 500 DMPs for each group (for cohorts with less than 500 DMPs, all DMPS were used). Cohort samples were aggregated using the median value of each probe within a group. A leaf node represents a cohort, with node sizes illustrating relative scales of the number of selected DMPs for the respective cohort, and node colors are indicative of the global mean methylation difference, a gradient of hypomethylation (blue) or hypermethylation (red). D Circular karyotype plot showing overlap of CHD2 450 K episignature probes (inner circle, n = 200), with CHD2 850 K episignature probes (middle circle, n = 200), and WGBS DMRs (n = 4 CHD2 vs. n = 6 unaffected controls) derived with at least a 15% methylation difference for the condensed visual representation (outer circle, n = 411). Each line depicts a probe or DMR where red denotes hypermethylation and blue denotes hypomethylation. The purple tracks depict coverage of the 450 K array probes (inner), 850 K EPIC array probes (middle), and WGBS reads (outer). Refer to Supplementary Fig. 33 for linear karyotype DMR plots for chr1-22.CC BY · PubMed Central
Fig. 7 The CHD2 Episignature is associated with DMRs enriched in regulatory regions. A Karyotype plots depict direct overlap of multiple CHD2 episignature probes with DMRs (left=hypermethylated region, right=hypomethylated region) called from WGBS (n = 4 CHD2 vs. n = 6 unaffected controls). For each karyotype plot, the three grey tracks (upper panel above chromosome) depict individual red (hyper) or blue (hypo) dots for WGBS DMRs (upper), CHD2 850 K episignature probes (middle) and CHD2 450 K episignature probes (lower). The scale denotes the methylation difference between CHD2 relative to controls. Three purple tracks (lower panel below chromosome) depict the coverage for the 450 K array probes (lines, upper), 850 K array probes (lines, middle), and WGBS reads (distribution, lower). The coverage track for the WGBS was taken from a representative sample after inspecting the average coverage values across all the samples. “Zoomed” in karyotype plots of the boxed regions of probes clustering around for HOXA4 (hyper) and ZNF577 (hypo) are shown above or below the karyotype plots. Gene annotations are noted as within 200 bp or 1,500 bp of the transcription start site (TSS200, TSS1500), the 5’ untranslated region (5’UTR), or gene body (Body). For both examples, multiple episignature probes map to DMRs. B Enrichment (values in Supplementary Data 10) of CHD2 episignature probes and DMRs (n = 4767 from array and WGBS, Supplementary Data 8) in various regulatory regions (bivalent regions, enhancers, and promoters) annotated with GREEN-DB99 compared to randomly generated genomic regions (Replicate-1, Replicate-2, and Replicate-3, Supplementary Data 9) of equal number (n = 4767 regions each) and varying, comparable sizes (50–3100 bp in length). Fisher’s Exact P values were calculated using two-sided Fisher’s Exact tests to determine the significance of enrichment. In all cases of CHD2 vs. Replicate-X, Fisher P < 2.2e-16. Counts of the regions annotated displaying enrichment in CHD2 are shown in the upper bar plot, and relative proportions are shown in the lower plot. Transcription factor binding sites (TFBS) annotation counts are plotted in C. (Fisher P < 2.2e−16). DNase sites annotation counts are plotted in D. (Fisher P < 2.2e−16).
Fig. 7 The CHD2 Episignature is associated with DMRs enriched in regulatory regions. A Karyotype plots depict direct overlap of multiple CHD2 episignature probes with DMRs (left=hypermethylated region, right=hypomethylated region) called from WGBS (n = 4 CHD2 vs. n = 6 unaffected controls). For each karyotype plot, the three grey tracks (upper panel above chromosome) depict individual red (hyper) or blue (hypo) dots for WGBS DMRs (upper), CHD2 850 K episignature probes (middle) and CHD2 450 K episignature probes (lower). The scale denotes the methylation difference between CHD2 relative to controls. Three purple tracks (lower panel below chromosome) depict the coverage for the 450 K array probes (lines, upper), 850 K array probes (lines, middle), and WGBS reads (distribution, lower). The coverage track for the WGBS was taken from a representative sample after inspecting the average coverage values across all the samples. “Zoomed” in karyotype plots of the boxed regions of probes clustering around for HOXA4 (hyper) and ZNF577 (hypo) are shown above or below the karyotype plots. Gene annotations are noted as within 200 bp or 1,500 bp of the transcription start site (TSS200, TSS1500), the 5’ untranslated region (5’UTR), or gene body (Body). For both examples, multiple episignature probes map to DMRs. B Enrichment (values in Supplementary Data 10) of CHD2 episignature probes and DMRs (n = 4767 from array and WGBS, Supplementary Data 8) in various regulatory regions (bivalent regions, enhancers, and promoters) annotated with GREEN-DB99 compared to randomly generated genomic regions (Replicate-1, Replicate-2, and Replicate-3, Supplementary Data 9) of equal number (n = 4767 regions each) and varying, comparable sizes (50–3100 bp in length). Fisher’s Exact P values were calculated using two-sided Fisher’s Exact tests to determine the significance of enrichment. In all cases of CHD2 vs. Replicate-X, Fisher P < 2.2e-16. Counts of the regions annotated displaying enrichment in CHD2 are shown in the upper bar plot, and relative proportions are shown in the lower plot. Transcription factor binding sites (TFBS) annotation counts are plotted in C. (Fisher P < 2.2e−16). DNase sites annotation counts are plotted in D. (Fisher P < 2.2e−16).CC BY · PubMed Central

Verification

Claims the paper makes

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

Free: it reads the open-access text and checks the first 5 claims that cite a source we can check. Create a free account to check every claim.

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 96 references

Cited by

See all 31

Metadata

Authors
LaFlamme, C. W., Rastin, C., Sengupta, S., Pennington, H. E., Russ-Hall, S. J., Schneider, A. L., Bonkowski, E. S., Almanza Fuerte, E. P., Allan, T. J., Zalusky, M. P.-G., Goffena, J., Gibson, S. B., Nyaga, D. M., Lieffering, N., Hebbar, M., Walker, E. V., Darnell, D., Olsen, S. R., Kolekar, P., Djekidel, M. N., Rosikiewicz, W., McConkey, H., Kerkhof, J., Levy, M. A., Relator, R., Lev, D., Lerman-Sagie, T., Park, K. L., Alders, M., Cappuccio, G., Chatron, N., Demain, L., Genevieve, D., Lesca, G., Roscioli, T., Sanlaville, D., Tedder, M. L., Gupta, S., Jones, E. A., Weisz-Hubshman, M., Ketkar, S., Dai, H., Worley, K. C., Rosenfeld, J. A., Chao, H.-T., Undiagnosed Diseases Network, Neale, G., Carvill, G. L., University of Washington Center for Rare Disease Research, Wang, Z., Berkovic, S. F., Sadleir, L. G., Miller, D. E., Scheffer, I. E., Sadikovic, B., & Mefford, H. C.
Affiliations
  • St. Jude Children's Research Hospital (US)
  • Western University (CA)
  • Rhodes College (US)
  • The University of Melbourne (AU)
  • Seattle Children's Hospital (US)
  • University of Otago (NZ)
  • London Health Sciences Centre (CA)
  • Wolfson Medical Center (IL)
  • University of Colorado Anschutz (US)
  • Amsterdam University Medical Centers (NL)
  • Telethon Institute Of Genetics And Medicine (IT)
  • Lyon 1 Université (FR)
  • St Mary's Hospital (GB)
  • UNSW Sydney (AU)
  • Greenwood Genetic Center (US)
  • Children's Hospital at Westmead (AU)
  • University of Manchester (GB)
  • Baylor College of Medicine (US)
  • Northwestern University (US)
Corresponding author
Mefford, Heather C.
Venue
Nature Communications
Published
2024-08-06
Volume
15
Issue
1
Pages
6524-6524
ISSN
2041-1723 (ISSN-L)
OpenAlex
W4401368338
Cited by
31
References
95
Citations by year
2026: 13 · 2025: 16 · 2024: 2
FWCI
9.07
Type
journal article
Language
en
Open access
Licence
Locations
  • Nature Communications (publishedVersion, cc-by)
  • medRxiv (acceptedVersion)
  • PubMed Central (publishedVersion, cc-by-nc-nd)
  • PubMed (publishedVersion)
  • PubMed Central (submittedVersion)
  • Digital Commons-TMC (Texas Medical Center) (submittedVersion)
  • RePEc: Research Papers in Economics (submittedVersion)
  • DOAJ (DOAJ: Directory of Open Access Journals) (submittedVersion, cc-by-sa)
  • Minerva Access (University of Melbourne) (submittedVersion)
  • Minerva Access (University of Melbourne) (submittedVersion, cc-by)
Funders
Citizens United for Research in Epilepsy
Topics
  • Health Sciences › Medicine › Clinical Neurology › Epilepsy research and treatment
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Molecular Biology › Epigenetics and DNA Methylation
  • Life Sciences › Biochemistry, Genetics and Molecular Biology › Genetics › Genomics and Rare Diseases
Keywords
DNA methylation, developmental and epileptic encephalopathy, pediatric epilepsy, CHD2, differentially methylated regions, genetic diagnosis, copy number variation, balanced translocation, pathogenic variants, methylation array, bisulfite sequencing, long-read sequencing

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

How to cite

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

This page
CoChat. (2026). Citation check: Diagnostic utility of DNA methylation analysis in genetically unsolved pediatric epilepsies and CHD2 episignature refinement. https://app.cochat.ai/papers/10.1038/s41467-024-50159-6
The paper
LaFlamme, C. W., Rastin, C., Sengupta, S., Pennington, H. E., Russ-Hall, S. J., Schneider, A. L., Bonkowski, E. S., Almanza Fuerte, E. P., Allan, T. J., Zalusky, M. P.-G., Goffena, J., Gibson, S. B., Nyaga, D. M., Lieffering, N., Hebbar, M., Walker, E. V., Darnell, D., Olsen, S. R., Kolekar, P., . . . Mefford, H. C. (2024). Diagnostic utility of DNA methylation analysis in genetically unsolved pediatric epilepsies and CHD2 episignature refinement. Nature Communications. https://doi.org/10.1038/s41467-024-50159-6

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.