CoChat

Electrosensory neural responses to natural electro-communication stimuli are distributed along a continuum

Open access

The DOI matches its Crossref record; 83 of the 83 references it cites were found in a registry or an index.

Abstract

Neural heterogeneities are seen ubiquitously within the brain and greatly complicate classification efforts. Here we tested whether the responses of an anatomically well-characterized sensory neuron population to natural stimuli could be used for functional classification. To do so, we recorded from pyramidal cells within the electrosensory lateral line lobe (ELL) of the weakly electric fish Apteronotus leptorhynchus in response to natural electro-communication stimuli as these cells can be anatomically classified into six different types. We then used two independent methodologies to functionally classify responses: one relies of reducing the dimensionality of a feature space while the other directly compares the responses themselves. Both methodologies gave rise to qualitatively similar results: while ON and OFF-type cells could easily be distinguished from one another, ELL pyramidal neuron responses are actually distributed along a continuum rather than forming distinct clusters due to heterogeneities. We discuss the implications of our results for neural coding and highlight some potential advantages.

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 · 10.1371/journal.pone.0175322

Cites

83 of 83 references verified0 citation errors · 0 not checkable

Statistics

Figures

Fig 1 Establishing a functional classification using naturalistic communication stimuli. A: There are two types of pyramidal neurons, On- (blue) and Off- (red) type, which can be distinguished anatomically by the presence and absence of basilar dendrites, respectively (top). On- and Off-type pyramidal cells can furthermore be subdivided into six classes: On and Off-type superficial (S) intermediate (I) and deep (D) types which each exhibits different sized apical dendritic trees. There is a strong negative correlation between the size of the apical dendritic tree and the baseline (i.e., in the absence of stimulation) firing rate (S1 Fig). The baseline firing rate is indicated by colour saturation as per the colour bar above the circuit diagram. At the circuit level (bottom) within the pyramidal cell layer (orange boarder) all neurons receive input from sensory afferents encoding the animal’s self-generated electric field. On-type cells receive direct inputs from these afferents whereas Off-type cells receive indirect input via local inhibitory interneurons. All neuron classes project to the midbrain torus Semicircularis (not pictured here) while only deep neurons project to praeminentialis dorsalis (Pd) which provides different degrees of inhibitory feedback to superficial and intermediate pyramidal neurons via the eminentia granularis pars posterior (EGP). B: The four chirp stimuli featured in this study are shown in dark grey. A 25 ms response window following chirp onset is also indicated by a light grey window for two On-type chirps (3π/2, π) and the two Off-type chirps (π/2, 0). The 5 Hz beat stimulus is shown in black. C: A stimulus waveform is played to an awake and behaving animal while recordings are obtained from pyramidal cells within the lateral segment (LS) of the ELL. Example recordings from one On-type and one Off-type neuron are shown in response to a 5 Hz beat. Spike waveforms identified using spike sorting software are indicated for each cell (blue and red). The spike times were used to generate raster plots and peristimulus time histograms (as seen below the experimental setup). Example cells have peak stimulus driven firing rates of 136 Hz (On-type) and 123 Hz (Off-type) and their responses to the beat are in anti-phase. The color gradient in the color bar (bottom) is indicative of the response magnitude of recorded units (i.e. On- or Off-type). The transition from blue to red reflects an increase in response magnitude as the logarithm in base 10 of the stimulus driven peak-firing rate. D: A priori it is unclear whether ELL pyramidal cells can be functionally classified based on their responses to natural communication signals alone. There are two hypotheses: 1. Responses form distinct clusters This is schematized by a heatmap of response magnitude showing distinct response profiles. Directly beneath a hierarchical agglomerative clustering algorithm applied to a pairwise distance matrix representing the above heatmap results in a dendrogram (green) which is clearly divisible into distinct groups (dashed red line). 2. Responses do not form distinct clusters and instead form a continuum. The response heat map as in 1 thus gives rise to one clear transition between On- and Off-type cells. In this case a hierarchical agglomerative clustering algorithm applied to a pairwise distance matrix representing the above heatmap results in a dendrogram (green) that is only divisible into two groups (dashed red line), each of which constitutes a continuum.
Fig 1 Establishing a functional classification using naturalistic communication stimuli. A: There are two types of pyramidal neurons, On- (blue) and Off- (red) type, which can be distinguished anatomically by the presence and absence of basilar dendrites, respectively (top). On- and Off-type pyramidal cells can furthermore be subdivided into six classes: On and Off-type superficial (S) intermediate (I) and deep (D) types which each exhibits different sized apical dendritic trees. There is a strong negative correlation between the size of the apical dendritic tree and the baseline (i.e., in the absence of stimulation) firing rate (S1 Fig). The baseline firing rate is indicated by colour saturation as per the colour bar above the circuit diagram. At the circuit level (bottom) within the pyramidal cell layer (orange boarder) all neurons receive input from sensory afferents encoding the animal’s self-generated electric field. On-type cells receive direct inputs from these afferents whereas Off-type cells receive indirect input via local inhibitory interneurons. All neuron classes project to the midbrain torus Semicircularis (not pictured here) while only deep neurons project to praeminentialis dorsalis (Pd) which provides different degrees of inhibitory feedback to superficial and intermediate pyramidal neurons via the eminentia granularis pars posterior (EGP). B: The four chirp stimuli featured in this study are shown in dark grey. A 25 ms response window following chirp onset is also indicated by a light grey window for two On-type chirps (3π/2, π) and the two Off-type chirps (π/2, 0). The 5 Hz beat stimulus is shown in black. C: A stimulus waveform is played to an awake and behaving animal while recordings are obtained from pyramidal cells within the lateral segment (LS) of the ELL. Example recordings from one On-type and one Off-type neuron are shown in response to a 5 Hz beat. Spike waveforms identified using spike sorting software are indicated for each cell (blue and red). The spike times were used to generate raster plots and peristimulus time histograms (as seen below the experimental setup). Example cells have peak stimulus driven firing rates of 136 Hz (On-type) and 123 Hz (Off-type) and their responses to the beat are in anti-phase. The color gradient in the color bar (bottom) is indicative of the response magnitude of recorded units (i.e. On- or Off-type). The transition from blue to red reflects an increase in response magnitude as the logarithm in base 10 of the stimulus driven peak-firing rate. D: A priori it is unclear whether ELL pyramidal cells can be functionally classified based on their responses to natural communication signals alone. There are two hypotheses: 1. Responses form distinct clusters This is schematized by a heatmap of response magnitude showing distinct response profiles. Directly beneath a hierarchical agglomerative clustering algorithm applied to a pairwise distance matrix representing the above heatmap results in a dendrogram (green) which is clearly divisible into distinct groups (dashed red line). 2. Responses do not form distinct clusters and instead form a continuum. The response heat map as in 1 thus gives rise to one clear transition between On- and Off-type cells. In this case a hierarchical agglomerative clustering algorithm applied to a pairwise distance matrix representing the above heatmap results in a dendrogram (green) that is only divisible into two groups (dashed red line), each of which constitutes a continuum.CC BY · PubMed Central
Fig 2 Responses of LS pyramidal cells to the beat. A: Peri-stimulus histograms (left) and cycle histograms (right) from six example On-type cells labeled according to phase of response to a 5Hz beat and baseline activity. Cells with higher baseline firing rates respond strongly to beats while those with lower baseline firing rates respond more weakly. Black arrows in the cycle histograms indicate the preferred phase and the length of the arrow gives the vector strength. Bin volume is indicated by values located at π/4 radians of each cycle histogram. Peak response magnitude values of example neurons are indicated by upward and downward pointing triangles on the colorbar (top) reflecting the logged stimulus driven firing rate. B: Same as in A but for six example Off-type neurons. C: Population distribution of response phase for all recordings in this study having a Z-stat ≥ 4. The histogram (bin size = π/6) reveals a bimodal distribution. Fitting the distribution with a Gaussian mixture model (black line) indicates an average on response at 1.08 radians and an average off response at 4.60 radians. The population (n = 74) is evenly divided into On- and Off-type neurons having mean vector strengths of 0.4175 ± SE 0.038 and 0.4226 ± SE 0.8664 respectively (panel inset). D: Linear regression models indicate a slight positive correlation of 0.445 exists between vector strength and baseline firing rate (p = 0.006) for On-type however no significant correlation exists for Off-type. E: No correlation exists between phase of response and baseline firing rate for either On-type or Off-type neurons as indicated by linear regression models. The rest is as in D.
Fig 2 Responses of LS pyramidal cells to the beat. A: Peri-stimulus histograms (left) and cycle histograms (right) from six example On-type cells labeled according to phase of response to a 5Hz beat and baseline activity. Cells with higher baseline firing rates respond strongly to beats while those with lower baseline firing rates respond more weakly. Black arrows in the cycle histograms indicate the preferred phase and the length of the arrow gives the vector strength. Bin volume is indicated by values located at π/4 radians of each cycle histogram. Peak response magnitude values of example neurons are indicated by upward and downward pointing triangles on the colorbar (top) reflecting the logged stimulus driven firing rate. B: Same as in A but for six example Off-type neurons. C: Population distribution of response phase for all recordings in this study having a Z-stat ≥ 4. The histogram (bin size = π/6) reveals a bimodal distribution. Fitting the distribution with a Gaussian mixture model (black line) indicates an average on response at 1.08 radians and an average off response at 4.60 radians. The population (n = 74) is evenly divided into On- and Off-type neurons having mean vector strengths of 0.4175 ± SE 0.038 and 0.4226 ± SE 0.8664 respectively (panel inset). D: Linear regression models indicate a slight positive correlation of 0.445 exists between vector strength and baseline firing rate (p = 0.006) for On-type however no significant correlation exists for Off-type. E: No correlation exists between phase of response and baseline firing rate for either On-type or Off-type neurons as indicated by linear regression models. The rest is as in D.CC BY · PubMed Central
Fig 3 Summary of steps taken in order to classify neuronal responses to naturalistic communication stimuli. The Common factor analysis technique (orange) aims to reduce dimensionality by developing a linear statistical model summarising in a low dimensional space the high dimensional response space of the data. Proximities within this space can then be used determine how responses are represented in the brain (green) (i.e. discrete clustered representations or a continuous representation). In contrast, the Dynamic time warping technique (blue) permits one to directly quantify the proximity between observations via a non-linear relation among responses abstracted as time series. Raster plots are transformed into time series (PSTHs). After defining a window of comparison to be permitted between PSTHs all pairwise comparisons between observations belonging to a population are made, yielding a pairwise distance matrix, which can then be used in the same way as for the Common Factor Analysis technique (green).
Fig 3 Summary of steps taken in order to classify neuronal responses to naturalistic communication stimuli. The Common factor analysis technique (orange) aims to reduce dimensionality by developing a linear statistical model summarising in a low dimensional space the high dimensional response space of the data. Proximities within this space can then be used determine how responses are represented in the brain (green) (i.e. discrete clustered representations or a continuous representation). In contrast, the Dynamic time warping technique (blue) permits one to directly quantify the proximity between observations via a non-linear relation among responses abstracted as time series. Raster plots are transformed into time series (PSTHs). After defining a window of comparison to be permitted between PSTHs all pairwise comparisons between observations belonging to a population are made, yielding a pairwise distance matrix, which can then be used in the same way as for the Common Factor Analysis technique (green).CC BY · PubMed Central
Fig 4 Pyramidal cell responses to chirp stimuli form a continuum based on an unsupervised classification algorithm including common factor analysis. A: Optimally sorted dendrogram (green) tracing the path of a single linkage agglomerative clustering algorithm from the leaves (right) to the root (left) as it was applied to the pairwise distance matrix representing Euclidian distance for all pairwise comparisons between observations projected into eight dimensions using a factor analytic model. The baseline firing rate is also indicated using the same color code as previously (see Fig 1A). B: Concatenated PSTH responses to the four chirp stimuli used in the study are presented for each neuron in the same order as the adjacent dendrogram. Responses are not normalized so that differences in response magnitude are readily apparent with the color gradient representing the logged stimulus driven firing rate allowing for a detailed visualization of each neuron’s stimulus preference. The baseline firing rate is also indicated using the same color code as previously.
Fig 4 Pyramidal cell responses to chirp stimuli form a continuum based on an unsupervised classification algorithm including common factor analysis. A: Optimally sorted dendrogram (green) tracing the path of a single linkage agglomerative clustering algorithm from the leaves (right) to the root (left) as it was applied to the pairwise distance matrix representing Euclidian distance for all pairwise comparisons between observations projected into eight dimensions using a factor analytic model. The baseline firing rate is also indicated using the same color code as previously (see Fig 1A). B: Concatenated PSTH responses to the four chirp stimuli used in the study are presented for each neuron in the same order as the adjacent dendrogram. Responses are not normalized so that differences in response magnitude are readily apparent with the color gradient representing the logged stimulus driven firing rate allowing for a detailed visualization of each neuron’s stimulus preference. The baseline firing rate is also indicated using the same color code as previously.CC BY · PubMed Central
Fig 5 Pyramidal cell responses to chirp stimuli form a continuum based on an unsupervised classification algorithm including dynamic time warping. A: Optimally sorted dendrogram (green). The color code is the same as in Fig 4A. B: Summarized concatenated PSTH response to the four chirp stimuli used in the study are presented for each neuron in the same order as the adjacent dendrogram. The color code is the same as in Fig 4B.
Fig 5 Pyramidal cell responses to chirp stimuli form a continuum based on an unsupervised classification algorithm including dynamic time warping. A: Optimally sorted dendrogram (green). The color code is the same as in Fig 4A. B: Summarized concatenated PSTH response to the four chirp stimuli used in the study are presented for each neuron in the same order as the adjacent dendrogram. The color code is the same as in Fig 4B.CC BY · PubMed Central
Fig 6 Responses of LS pyramidal cells to chirps. A: Illustration of the methodology used to differentiate between the responses to the beat and to the chirp. The chirp stimuli of interest are shown in green and the corresponding responses from typical On-type (blue) and Off-type (red) neurons are also shown running the full extent of the stimulus. The response to a beat stimulus is then aligned in phase with the beat of the stimulus of interest both before and after the chirp. These two alignments are indicated by two separate dashed lines identified as the pre-chirp and post-chirp beat and run the full extent of the stimulus of interest. Directly beneath actual responses is a signal which can take on both positive and negative values as it was generated by subtracting the pre-chirp and post-chirp responses from the response to the stimulus of interest. The line running though or above this signal indicates a value of zero with positive values highlighted an appropriate color. The maximum value of this signal within the grey window (25 msec after chirp onset) is taken as the response of the neuron to the chirp. Responses to each of the four chirps are used to generate a 2 dimensional representation of the 4 dimensional response space known as a glyph. The correspondence between glyph dimensions and neural response to chirp phases are demonstrated for average On- and Off -type examples. Correspondence is indicated by highlighting the glyph axis associated with a given chirp phase on the glyph seen to the right of that chirp phase response. B: Peri-stimulus histograms from the same six example On-type cells used in Fig 2. Responses to the 4 different chirps were concatenated. Note that, while responses of superficial On-type cells to the beat are difficult to discern from the PSTH’s in Fig 2, their responses to chirps are quite clear. A glyph summarizing each example neuron’s location within the response space to these four chirps is located to the right of their PSTH and their logged peak firing rate response is indicated by a leftward or rightward pointing triangle on the adjacent colorbar. C: Same as in B but for 6 example Off-type neurons. D: Representation of the response space to 4 natural communication signals averaging over different populations. (Top left) Chirp responses of all On-type cells were averaged along each dimension of response space to generate an average “On glyph”. The same was done for all Off-type cells. Multidimensional scaling was used to project the response space into two dimensions and glyphs where plotted centered on their two coordinate representation. The visualization procedure was repeated but for more specific subpopulations by dividing On- and Off-type further into deep, intermediate, and superficial (bottom left). For comparison, the glyphs from individual neurons are also shown (right).
Fig 6 Responses of LS pyramidal cells to chirps. A: Illustration of the methodology used to differentiate between the responses to the beat and to the chirp. The chirp stimuli of interest are shown in green and the corresponding responses from typical On-type (blue) and Off-type (red) neurons are also shown running the full extent of the stimulus. The response to a beat stimulus is then aligned in phase with the beat of the stimulus of interest both before and after the chirp. These two alignments are indicated by two separate dashed lines identified as the pre-chirp and post-chirp beat and run the full extent of the stimulus of interest. Directly beneath actual responses is a signal which can take on both positive and negative values as it was generated by subtracting the pre-chirp and post-chirp responses from the response to the stimulus of interest. The line running though or above this signal indicates a value of zero with positive values highlighted an appropriate color. The maximum value of this signal within the grey window (25 msec after chirp onset) is taken as the response of the neuron to the chirp. Responses to each of the four chirps are used to generate a 2 dimensional representation of the 4 dimensional response space known as a glyph. The correspondence between glyph dimensions and neural response to chirp phases are demonstrated for average On- and Off -type examples. Correspondence is indicated by highlighting the glyph axis associated with a given chirp phase on the glyph seen to the right of that chirp phase response. B: Peri-stimulus histograms from the same six example On-type cells used in Fig 2. Responses to the 4 different chirps were concatenated. Note that, while responses of superficial On-type cells to the beat are difficult to discern from the PSTH’s in Fig 2, their responses to chirps are quite clear. A glyph summarizing each example neuron’s location within the response space to these four chirps is located to the right of their PSTH and their logged peak firing rate response is indicated by a leftward or rightward pointing triangle on the adjacent colorbar. C: Same as in B but for 6 example Off-type neurons. D: Representation of the response space to 4 natural communication signals averaging over different populations. (Top left) Chirp responses of all On-type cells were averaged along each dimension of response space to generate an average “On glyph”. The same was done for all Off-type cells. Multidimensional scaling was used to project the response space into two dimensions and glyphs where plotted centered on their two coordinate representation. The visualization procedure was repeated but for more specific subpopulations by dividing On- and Off-type further into deep, intermediate, and superficial (bottom left). For comparison, the glyphs from individual neurons are also shown (right).CC BY · PubMed Central
Fig 7 A: At each step of the single linkage algorithm clusters are merged on the basis of the minimum distance between two observations each belonging to separate clusters. Which observations were linked at each iteration of the algorithm (nodes of the above dendrogram) and at what distance they were from each other within the original pairwise distance matrix was used to generate an adjacency matrix or network that was then represented as a network graph. At each two dimensional coordinate the glyph summarizing the observations location within the response space to 4 chirp stimuli was plotted and a colored letter indicating On- (blue) or Off-type (red), deep (D), intermediate (I) or superficial (S) was plotted atop this. B: Same as A but for DTW. C: Response profiles captured by 44 response features projected into a three dimensional feature space using an 8 Factor statistical model (χ2 = 1266 d.f. = 622 p = 3.47x10-46) accounted for 78% of the variance. Each observation is colored according to its anatomical designation and each observation is connected to the mean value of its anatomical class within the factor space. D: Network graph (desaturated black lines) constructed from an adjacency matrix where each observation was connected to the closest one half of the population using the warping distance between observations as entries with each mode labeled according to On- or Off-type as well as anatomical cell class. The mean coordinate for each of the six anatomical designations was computed and observations were linked to their respective groups mean.
Fig 7 A: At each step of the single linkage algorithm clusters are merged on the basis of the minimum distance between two observations each belonging to separate clusters. Which observations were linked at each iteration of the algorithm (nodes of the above dendrogram) and at what distance they were from each other within the original pairwise distance matrix was used to generate an adjacency matrix or network that was then represented as a network graph. At each two dimensional coordinate the glyph summarizing the observations location within the response space to 4 chirp stimuli was plotted and a colored letter indicating On- (blue) or Off-type (red), deep (D), intermediate (I) or superficial (S) was plotted atop this. B: Same as A but for DTW. C: Response profiles captured by 44 response features projected into a three dimensional feature space using an 8 Factor statistical model (χ2 = 1266 d.f. = 622 p = 3.47x10-46) accounted for 78% of the variance. Each observation is colored according to its anatomical designation and each observation is connected to the mean value of its anatomical class within the factor space. D: Network graph (desaturated black lines) constructed from an adjacency matrix where each observation was connected to the closest one half of the population using the warping distance between observations as entries with each mode labeled according to On- or Off-type as well as anatomical cell class. The mean coordinate for each of the six anatomical designations was computed and observations were linked to their respective groups mean.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 83 references

Research works printed but not linked: 2

  1. Metzen MG, Hofmann V, Chacron MJ. Neural correlations enable invariant coding and perception of natural stimuli in weakly electric fish. Elife. 2016; 5: e12993 10.7554/eLife.12993 27128376 PMC4851552 · DOI
  2. Rath TM, Manmatha R, Ieee Computer S, Ieee Computer S. Word image matching using dynamic time warping. 2003 Ieee Computer Society Conference on Computer Vision and Pattern Recognition, Vol Ii, Proceedings. Proceedings—Ieee Computer Society Conference on Computer Vision and Pattern Recognition. Los Alamitos: Ieee Computer Soc; 2003. p. 521–7.

Cited by

See all 9

Metadata

Authors
Sproule, M. K. J., & Chacron, M. J.
Affiliations
McGill University (CA)
Corresponding author
  • Sproule, Michael K. J.
  • Chacron, Maurice J.
Venue
PLoS ONE
Published
2017-04-06
Volume
12
Issue
4
Pages
e0175322
ISSN
1932-6203 (ISSN-L)
OpenAlex
W2604524399
Cited by
9
References
83
Citations by year
2026: 1 · 2023: 2 · 2022: 1 · 2021: 1 · 2020: 1 · 2019: 1 · 2018: 2
FWCI
0.91
Type
journal article
Language
en
Open access
Locations
  • PLoS ONE (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)
  • Figshare (submittedVersion, cc-by)
Funders
  • Natural Sciences and Engineering Research Council of Canada
  • Canada Research Chairs
Topics
  • Physical Sciences › Environmental Science › Nature and Landscape Conservation › Fish biology, ecology, and behavior
  • Life Sciences › Neuroscience › Cognitive Neuroscience › Neural dynamics and brain function
  • Life Sciences › Neuroscience › Cellular and Molecular Neuroscience › Neuroscience and Neuropharmacology Research
Keywords
electrosensory lateral line lobe, Apteronotus leptorhynchus, weakly electric fish, natural stimuli, pyramidal neurons, neuronal heterogeneity, functional classification, neural coding, ON and OFF responses, dimensionality reduction
MeSH terms
Action Potentials, Animals, Cell Communication, Electric Fish, Electric Stimulation, Pyramidal Cells

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: Electrosensory neural responses to natural electro-communication stimuli are distributed along a continuum. https://app.cochat.ai/papers/10.1371/journal.pone.0175322
The paper
Sproule, M. K. J., & Chacron, M. J. (2017). Electrosensory neural responses to natural electro-communication stimuli are distributed along a continuum. PLoS ONE. https://doi.org/10.1371/journal.pone.0175322

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.