OSCR

A pipeline for developing AI-driven models to predict molecular initiating events: a case study on neural tube defects.

Code ↔ Paper

16 matches between paragraphs of the paper and lines of its authors' code, computed by the harvester (lexical-v1). Click a colored paragraph or line to see its counterpart.

The 16 matches
  1. [1] § Methods › Training and optimization › Hyper-parameter tuning ↔ scripts/experiment.py, lines 54–116 · score 0.82 · ASHAScheduler, Hyper parameters, stratified cross validation, weight decay, epochs, dropout
  2. [2] § Methods › Automatic curation pipeline ↔ scripts/autocurate.py, lines 134–159 · score 0.81 · canonical tautomeric, largest fragment, kekul SMILES, stereochemistry, kekulized, invalid
  3. [3] § Methods › Automatic curation pipeline ↔ scripts/autocurate.py, lines 46–125 · score 0.77 · numeric, endpoints, potency, pChEMBL, inhibition, flagged
  4. [4] § Results › A robust pipeline for automatic development of MIE prediction models ↔ scripts/run_pipeline.sh, lines 30–104 · score 0.75 · autocurate.py, experiment.py, prepare_molecular_graphs.py, ChEMBL, splits, pipeline
  5. [5] § Methods › Automatic curation pipeline ↔ scripts/data_analysis_scripts/compare_smiles.py, lines 32–49 · score 0.70 · largest fragment, kekul SMILES, stereochemistry, kekulized, canonical
  6. [6] § Results › A robust pipeline for automatic development of MIE prediction models ↔ scripts/autocurate.py, lines 1–21 · score 0.67 · RDKit, cleaning SMILES, ChEMBL, autocurate, duplicate, conflicting
  7. [7] § Methods › Ablation studies ↔ scripts/experiment.py, lines 54–116 · score 0.66 · KPGT OS, loss weighting, pre training, fine tuning, model
  8. [8] § Results › Predicting HDAC inhibitors › Ablation studies: identifying optimal KPGT configuration for pipeline integration ↔ scripts/data_analysis_scripts/analysis.py, lines 182–249 · score 0.65 · Wilcoxon Signed Rank, Bonferroni correction, ablations, folds, metrics, MCC
  9. [9] § Methods › Automatic curation pipeline ↔ scripts/autocurate.py, lines 178–258 · score 0.62 · ChEMBL ID, provenance, Excel, minimal, raw, curated
  10. [10] § Results › Predicting HDAC inhibitors › Proof-of-concept: KPGT outperforms the benchmark SVM-RBF ↔ src/trainer/evaluator.py, lines 226–263 · score 0.60 · MCC score, balanced accuracy, sensitivity, TN, TP, SPE
  11. [11] § Methods › Automatic curation pipeline ↔ scripts/autocurate.py, lines 178–258 · score 0.54 · pChEMBL, OpenPyXL, cutoff, threshold, Python, activity
  12. [12] § Methods › Evaluation metrics ↔ src/trainer/evaluator.py, lines 226–263 · score 0.54 · balanced accuracy, sensitivity, TN, TP, SPE, FN
  13. [13] § Methods › Training and optimization › Hyper-parameter tuning ↔ scripts/finetune.py, lines 143–292 · score 0.53 · weight decay, cross validation, epochs, optimized, dropout, split
  14. [14] § Methods › Automatic curation pipeline ↔ scripts/finetune.py, lines 143–292 · score 0.53 · inference_meta.json, trained model, configuration, molecular, Predictions
  15. [15] § Results › Predicting HDAC inhibitors › Proof-of-concept: KPGT outperforms the benchmark SVM-RBF ↔ scripts/data_analysis_scripts/analysis.py, lines 252–284 · score 0.52 · Post hoc, Friedman, Nemenyi, ANOVA, folds, models
  16. [16] § Methods › Automatic curation pipeline ↔ scripts/manual/preprocess_manual_curated.py, lines 52–71 · score 0.50 · ChEMBL IDs, inactive, curated, class, SMILES

Paper

Loaded from Europe PMC by your browser, not stored by OSCR: doi.org · Europe PMC

The paper is loaded when this pane is shown.

The authors' code

Python · 262 lines · 11 KB · Apache-2.0 · 5 matches

  1. #!/usr/bin/env python
  2. """
  3. build_dataset.py
  4. ================
  5. raw ChEMBL dump → curated (activity rules) → clean SMILES
  6. → drop conflicts / duplicates → write 3‑column CSV
  8. Requires RDKit ≥ 2022.03 (MolStandardize), pandas.
  9. """
  11. from __future__ import annotations
  12. import argparse, re
  13. from pathlib import Path
  15. import pandas as pd
  16. from rdkit import Chem
  17. from rdkit.Chem import MolStandardize # for older RDKit
  18. try: # RDKit ≥ 2023.09
  19. from rdkit.Chem.MolStandardize import rdMolStandardize as rdms
  20. except ImportError: # RDKit ≤ 2023.03
  21. from rdkit.Chem import rdMolStandardize as rdms
  24. # ────────────────────────────── 1. argument parser ────────────────────── #
  26. def get_args() -> argparse.Namespace:
  27. p = argparse.ArgumentParser()
  28. p.add_argument("--input", required=True, help="raw ChEMBL .csv/.tsv")
  29. p.add_argument("--dataset", required=True, help="folder / base‑name")
  30. p.add_argument("--outdir", default="../datasets", help="root output dir")
  31. p.add_argument("--sheet", default="curated",
  32. help="sheet name for the cleaned Excel copy")
  33. p.add_argument("--pchembl_thresh", type=float, default=7.0,
  34. help="if Category missing: pChEMBL ≥ thresh ⇒ active")
  35. p.add_argument("--cutoff_nm", type=float, default=10_000,
  36. help="activity cut‑off in nM (default 10 µM)")
  37. return p.parse_args()
  40. # ───────────────────────────── 2. activity curation ───────────────────── #
  42. PCHEMBL_TYPES = ["IC50", "XC50", "EC50", "AC50", "Ki", "Kd",
  43. "Potency", "ED50"]
  44. EXCLUDE_WORDS = ("inconclusive", "undetermined", "not determined")
  46. def curate_activity(df: pd.DataFrame, cutoff_nm: float) -> pd.DataFrame:
  47. """Numerical + text rule‑set to produce a binary Category column."""
  48. out = df.copy()
  50. # 2.1 keep potency‑like endpoints
  51. out = out[out["Standard Type"].isin(PCHEMBL_TYPES)]
  53. # 2.2 drop rows flagged by ChEMBL
  54. out = out[out["Data Validity Comment"].isna()]
  56. # 2.3 free‑text veto
  57. pattern = "|".join(map(re.escape, EXCLUDE_WORDS))
  58. out["Comment"] = out["Comment"].fillna("")
  59. out = out[~out["Comment"].str.contains(pattern, case=False)]
  61. # 2.4 col clean‑up
  62. out["Standard Value"] = pd.to_numeric(out["Standard Value"],
  63. errors="coerce")
  64. out["Standard Relation"] = (out["Standard Relation"]
  65. .astype(str).str.strip())
  66. out["Standard Units"] = out["Standard Units"].astype(str).str.strip()
  67. out["Standard Relation"] = \
  68. out["Standard Relation"].str.extract(r"([<>=]+)")\
  69. .fillna(out["Standard Relation"])
  71. # 2.5 binary label
  72. def classify(row):
  73. sv = row["Standard Value"]
  74. rel = str(row["Standard Relation"]).strip()
  75. unit = str(row["Standard Units"]).strip().lower()
  77. # ── 1. relation‑driven decisions ─────────────────────────────
  78. # “greater‑than” ⇒ always not active, even if sv is NaN
  79. if rel in (">", ">="):
  80. return "not active"
  82. # “less‑than / equal / empty” needs a number to judge
  83. if rel in ("<", "<=", "=","") and pd.notna(sv) and unit == "nm":
  84. return "active" if sv < cutoff_nm else "not active"
  86. # ── 2. fall back to free‑text comments ──────────────────────
  87. comment = (row["Comment"] or "").lower()
  88. if any(w in comment for w in ("not active", "no inhibition", "inactive")):
  89. return "not active"
  90. if any(w in comment for w in ("active", "inhibitor")):
  91. return "active"
  93. # ── 3. could not decide ─────────────────────────────────────
  94. return "unknown"
  95. # 2.5 binary label
  97. # def classify(row):
  98. # sv, rel, unit = row["Standard Value"], row["Standard Relation"], row["Standard Units"]
  99. # if pd.notna(sv) and unit == "nM":
  100. # if rel in ("<", "<=", "="):
  101. # return "active" if sv < cutoff_nm else "not active"
  102. # if rel in (">", ">="): # treat any “greater than” as inactive
  103. # return "not active"
  104. # # fallback: comments
  105. # text = row["Comment"].lower()
  106. # if "active" in text or "inhibitor" in text: return "active"
  107. # if "not active" in text or "no inhibition" in text: return "not active"
  108. # return "unknown"
  110. out["Category"] = out.apply(classify, axis=1)
  111. n_unknown = (out["Category"] == "unknown").sum()
  112. out = out[out["Category"] != "unknown"]
  114. # 2.6 reliability tag (not used downstream but good to keep)
  115. def reliability(row):
  116. if pd.notna(row["Standard Relation"]) and pd.notna(row["Standard Value"]):
  117. return "high"
  118. if pd.notna(row["Standard Value"]): return "moderate"
  119. if pd.notna(row["Comment"]): return "moderate"
  120. return "low"
  122. out["Category (reliability)"] = out.apply(reliability, axis=1)
  123. if n_unknown:
  124. print(f"• dropped {n_unknown} rows with un‑classifiable activity")
  125. return out.reset_index(drop=True)
  128. # ─────────────────────────── 3. SMILES cleaning ─────────────────────── #
  130. chooser = rdms.LargestFragmentChooser()
  131. uncharger = rdms.Uncharger()
  132. taut_enum = rdms.TautomerEnumerator()
  134. def clean_smiles(raw: str):
  135. """
  136. largest fragment → uncharged → canonical tautomer
  137. → no stereo → kekulise → canonical SMILES
  138. returns (smiles, tag)
  139. """
  140. try:
  141. mol = Chem.MolFromSmiles(raw)
  142. if mol is None:
  143. raise ValueError("invalid SMILES")
  145. mol = chooser.choose(mol)
  146. mol = uncharger.uncharge(mol)
  147. mol = taut_enum.Canonicalize(mol)
  149. Chem.SanitizeMol(mol, catchErrors=True)
  150. Chem.RemoveStereochemistry(mol)
  151. Chem.Kekulize(mol, clearAromaticFlags=True)
  153. smi = Chem.MolToSmiles(mol,
  154. kekuleSmiles=True,
  155. canonical=True)
  156. return smi, "Clean"
  158. except Exception as exc:
  159. return None, f"Error: {exc}"
  162. # ─────────────────────── 4. duplicate / conflict pruning ────────────── #
  164. def prune_duplicates(df: pd.DataFrame) -> pd.DataFrame:
  165. """Drop rows where the same SMILES (or chembl_id) occurs with
  166. conflicting class labels, then de‑duplicate."""
  167. for col in ["smiles", "chembl_id"]:
  168. cross = (df.groupby([col, "Class"]).size()
  169. .unstack(fill_value=0))
  170. conflicts = cross[(cross["0"] > 0) & (cross["1"] > 0)].index
  171. df = (df[~df[col].isin(conflicts)]
  172. .drop_duplicates(subset=col, keep="first"))
  173. return df
  176. # ────────────────────────────── 5. main ─────────────────────────────── #
  178. def main() -> None:
  179. args = get_args()
  181. in_path = Path(args.input)
  182. base_dir = (Path(args.outdir).expanduser() / args.dataset)
  183. base_dir.mkdir(parents=True, exist_ok=True)
  185. # 5.1 read & curate
  186. raw = pd.read_csv(in_path, sep=None, engine="python")
  187. curated = curate_activity(raw, args.cutoff_nm)
  189. # 5.2 add row‑counter (for provenance only)
  190. curated.insert(0, "row_id", range(1, len(curated)+1))
  192. # 5.3 SMILES cleaning
  193. keku, status = zip(*curated["Smiles"].map(clean_smiles))
  194. curated["SMILES (Kekulized)"] = keku
  195. curated["Cleaning Status"] = status
  197. n_fail = curated["SMILES (Kekulized)"].isna().sum()
  198. if n_fail:
  199. print(f"• dropped {n_fail} rows whose SMILES failed to clean")
  200. curated = curated[curated["SMILES (Kekulized)"].notna()]
  202. # 5.4 fallback Category from pChEMBL (if missing)
  203. if "Category" not in curated.columns:
  204. curated["Category"] = curated["pChEMBL Value"].apply(
  205. lambda x: "active" if pd.notna(x) and float(x) >= args.pchembl_thresh
  206. else "not active")
  207. curated["Category"] = curated["Category"].str.strip().str.lower()
  209. # Drop any SMILES that appears with *both* labels
  210. conflict_smiles = (curated.groupby("SMILES (Kekulized)")["Category"]
  211. .nunique()
  212. .loc[lambda s: s > 1] # keeps only >1‑label cases
  213. .index)
  215. n_conf = curated["SMILES (Kekulized)"].isin(conflict_smiles).sum()
  216. if n_conf:
  217. print(f"• removed {n_conf} rows with conflicting activity labels")
  218. curated = curated[~curated["SMILES (Kekulized)"].isin(conflict_smiles)]
  220. # 5.5 save full Excel for provenance
  221. xlsx_path = base_dir / f"{args.dataset}.xlsx"
  222. with pd.ExcelWriter(xlsx_path, engine="openpyxl") as xl:
  223. raw.to_excel(xl, sheet_name="raw", index=False)
  224. curated.to_excel(xl, sheet_name=args.sheet, index=False)
  226. # ───────── minimal 3‑column frame for ML ─────────────────────────── #
  228. df = (curated[["SMILES (Kekulized)", "Category", "Molecule ChEMBL ID"]]
  229. .rename(columns={"SMILES (Kekulized)": "smiles",
  230. "Category": "Class",
  231. "Molecule ChEMBL ID": "chembl_id"}))
  233. df["Class"] = df["Class"].map({"active": "1", "not active": "0"})
  234. df = df.sort_values("Class", ascending=False)
  236. before = len(df)
  237. df = prune_duplicates(df)
  238. pruned = before - len(df)
  239. if pruned:
  240. print(f"• removed {pruned} duplicate / conflicting rows")
  242. # 5.6 final stats & write files
  243. n_tot = len(df)
  244. n_act = (df["Class"] == "1").sum()
  245. n_inact = n_tot - n_act
  246. print(f"✓ final dataset: {n_tot} rows "
  247. f"({n_act} active / {n_inact} inactive)")
  249. csv_id = base_dir / f"{args.dataset}_with_id.csv"
  250. csv_main = base_dir / f"{args.dataset}.csv"
  251. df.to_csv(csv_id, index=False) # with chembl_id
  252. df.drop(columns="chembl_id").to_csv(csv_main, index=False)
  254. print("Files written:")
  255. for p in (xlsx_path, csv_id, csv_main):
  256. try: print(" ", p.relative_to(Path.cwd()))
  257. except ValueError:
  258. print(" ", p)
  261. if __name__ == "__main__":
  262. main()

autocurate.py at commit b2b5668, under Apache-2.0 · at the source

Overview

Authors: Job H Berkhout1,2, Merel Florian1,3, Domenico Gadaleta4, Aldert H Piersma1,2, Harm J Heusinkveld1
  1. Centre for Health Protection, National Institute for Public Health and the Environment, Bilthoven, The Netherlands
  2. Institute for Risk Assessment Sciences, Utrecht University, Utrecht, The Netherlands
  3. Learning and Reasoning, Vrije Universiteit Amsterdam, Amsterdam, The Netherlands
  4. Laboratory of Environmental Chemistry and Toxicology, Istituto di Ricerche Farmacologiche Mario Negri IRCCS, Milan, Italy
Institutions: Utrecht University (Netherlands); National Institute for Public Health and the Environment (Netherlands); Vrije Universiteit Amsterdam (Netherlands); Mario Negri Institute for Pharmacological Research (Italy)
Journal: Journal of cheminformatics, volume 18, issue 1, article 62
Dates: received 13 November 2025; accepted 26 February 2026; published online 2 April 2026
Type: Research article · Language: English
License: CC BY-NC-ND
Identifiers: DOI 10.1186/s13321-026-01177-7 · PMID 41928299 · PMCID PMC13169617 · OpenAlex W7148596742
Open access: gold, a free copy (OpenAlex)
Status: code verified
Methods: Machine learning, Statistics, Graphs
Keywords: Computational toxicology, Deep Learning, Adverse Outcome Pathway, Neural Tube Closure
Topic: Machine Learning in Materials Science (Materials Chemistry, Materials Science), according to OpenAlex
Funding: Horizon 2020 Framework Programme (No 963845)
Citations: not cited yet (Europe PMC); 70 references in the paper

Abstract

The abstract is not reproduced here: the paper's license (CC BY-NC-ND) does not allow it. Read it in the paper, at the publisher or on Europe PMC.

Repository

Its files are read in the Code ↔ Paper reader above, with 16 matches between paragraphs and lines of code.

MerelFlorian/NeuroTox-KPGT

License: Apache-2.0
State: the link answers, verified on 28 September 2026
Evidence: files inventoried
Commit: b2b566837fb99210f2473cff7cde529e45df4a02, 12 November 2025
Languages: Python (45), Shell (1)
Size: 80 files, 46 scripts
Software Heritage: not archived
Found in: the text, “Automatic curation pipeline”
Holds: README, license file, environment (environment_exp.yml)
Not found: CITATION.cff, tests, continuous integration, documentation
Tools: NumPy (29 files), PyTorch (19 files), pandas (15 files), RDKit (9 files), SciPy (7 files), scikit-learn (6 files), Matplotlib (3 files), NetworkX (2 files), seaborn (2 files), Pingouin (1 file), scikit-posthocs (1 file), statsmodels (1 file)
Availability: 1 check, the latest on 28 September 2026: the link answers
  • 28 September 2026: the link answers
48 files
At the source: github.com/MerelFlorian/NeuroTox-KPGT.git

The paper's code and data availability statement is in the Data section.

Tracing map

Proposed by the machine: these links were found in the paper and verified at the source, without human review. The map will receive a Zenodo DOI once one of the paper's authors has validated it with their ORCID.

What the map holds:

  • 1 repository of the authors' code, each at its verified commit, with its license and how the link was found in the paper;
  • 46 scripts, each with its path and the digest of its content;
  • 16 matches between paragraphs of the paper and lines of the code (method lexical-v1);
  • neither the text of the paper nor the code itself.

Its JSON (tracing-map.json) is deposited on Zenodo with its DOI once the map is validated.

Data

No dataset and no data link were found in the paper.

Code and data availability statement

The paper has a code and data availability statement. Its license (CC BY-NC-ND) does not allow reproducing it here; in short, from what the harvester recognized in it:

  • no repository, dataset or request procedure was recognized in it

Read it in the paper: doi.org/10.1186/s13321-026-01177-7.

Versions

The history of this record: each version stored by the harvester or made by a correction of its authors or of the maintainers of its code, and what changed in its facts. The texts of the paper (its abstract, its availability statements) are not part of it; versions that changed only those are not listed.

Version 1, 28 September 2026: the first record

Recorded: type, language, journal, volume, issue, pages, dates, 5 authors, 4 keywords, 1 funder, 38 references.

Cite

This paper

Berkhout, J. H., Florian, M., Gadaleta, D., H Piersma, A., & Heusinkveld, H. J. (2026). A pipeline for developing AI-driven models to predict molecular initiating events: a case study on neural tube defects. Journal of cheminformatics, 18(1), 62. https://doi.org/10.1186/s13321-026-01177-7

BibTeX

@article{berkhout2026pipeline,
author = {Berkhout, Job H and Florian, Merel and Gadaleta, Domenico and H Piersma, Aldert and Heusinkveld, Harm J},
title = {{A pipeline for developing AI-driven models to predict molecular initiating events: a case study on neural tube defects}},
journal = {Journal of cheminformatics},
year = {2026},
month = apr,
volume = {18},
number = {1},
pages = {62},
publisher = {BMC},
issn = {1758-2946},
doi = {10.1186/s13321-026-01177-7},
url = {https://doi.org/10.1186/s13321-026-01177-7},
pmid = {41928299},
pmcid = {PMC13169617}
}

RIS

TY - JOUR
AU - Berkhout, Job H
AU - Florian, Merel
AU - Gadaleta, Domenico
AU - H Piersma, Aldert
AU - Heusinkveld, Harm J
TI - A pipeline for developing AI-driven models to predict molecular initiating events: a case study on neural tube defects
T2 - Journal of cheminformatics
J2 - J Cheminform
PY - 2026
DA - 2026/04/02
VL - 18
IS - 1
SP - 62
SN - 1758-2946
PB - BMC
DO - 10.1186/s13321-026-01177-7
UR - https://doi.org/10.1186/s13321-026-01177-7
LA - en
ER -

CSL-JSON

{
"id": "10.1186/s13321-026-01177-7",
"type": "article-journal",
"title": "A pipeline for developing AI-driven models to predict molecular initiating events: a case study on neural tube defects",
"container-title": "Journal of cheminformatics",
"author": [
{
"family": "Berkhout",
"given": "Job H"
},
{
"family": "Florian",
"given": "Merel"
},
{
"family": "Gadaleta",
"given": "Domenico"
},
{
"family": "H Piersma",
"given": "Aldert"
},
{
"family": "Heusinkveld",
"given": "Harm J"
}
],
"container-title-short": "J Cheminform",
"volume": "18",
"issue": "1",
"page": "62",
"DOI": "10.1186/s13321-026-01177-7",
"PMID": "41928299",
"PMCID": "PMC13169617",
"ISSN": "1758-2946",
"publisher": "BMC",
"URL": "https://doi.org/10.1186/s13321-026-01177-7",
"language": "en",
"issued": {
"date-parts": [
[
2026,
4,
2
]
]
}
}

The tracing map gets a citation of its own once an author has validated it and it has a DOI.

Similar papers

The papers with a page that share the most with this one: the tools found in their code, their categories, datasets, cited references and authors, the rarest counting most.

[1] doi:10.1016/j.isci.2026.117375 [code]
Motor priming is associated with widespread recruitment into neural ensembles and more rapid ensemble transitions.
Journal: iScience
In common: scikit-posthocs, Pingouin, NetworkX, 8 other tools
[2] doi:10.1038/s42003-026-10957-8 [code]
Brain defence by the extracellular matrix protein Cochlin.
Journal: Communications biology
In common: RDKit, NetworkX, statsmodels, 7 other tools
[3] doi:10.1093/nar/gkag706 [code]
scDifformer: diffusion-based post-training for virtual cell modeling across large-scale single-cell data.
Journal: Nucleic acids research
In common: RDKit, NetworkX, statsmodels, 7 other tools
[4] doi:10.1371/journal.pcbi.1014571 [code]
SynAPSeg: A novel dataset and image analysis framework for deep learning-based synapse detection and quantification.
Journal: PLoS computational biology
In common: Pingouin, NetworkX, statsmodels, 7 other tools
[5] doi:10.1016/j.isci.2026.116825 [code]
Social hierarchy shapes behavioral and transcriptional responses to chronic stress and ketamine in male mice.
Journal: iScience
In common: Pingouin, NetworkX, statsmodels, 7 other tools
[6] doi:10.1002/ana.78203 [code]
AI-Driven Mapping of Seizure Spread Patterns.
Journal: Annals of neurology
In common: Pingouin, NetworkX, statsmodels, 7 other tools
[7] doi:10.1038/s41593-026-02232-0 [code]
Entorhinal cortex represents task-relevant remote locations independently of CA1.
Journal: Nature neuroscience
In common: Pingouin, NetworkX, statsmodels, 7 other tools
[8] doi:10.1038/s41593-026-02362-5 [code]
Replay of procedural memory is independent of the hippocampus.
Journal: Nature neuroscience
In common: scikit-posthocs, Pingouin, statsmodels, 6 other tools
[9] doi:10.1038/s41467-026-74823-1 [code]
Cerebellar activity is triggered by reach endpoint during learning of a complex locomotor task.
Journal: Nature communications
In common: scikit-posthocs, Pingouin, statsmodels, 6 other tools
[10] doi:10.1038/s41467-026-75662-w [code]
Distinct Roles of Deep and Superficial Cortical Layers in Tone Prediction, Comparison, and Adaptation in Human Auditory Cortices.
Journal: Nature communications
In common: scikit-posthocs, Pingouin, statsmodels, 6 other tools

Contribute

The authors of this paper can claim it, correct its record and validate its tracing map, and the maintainers of its code (its owner, or a public member of its organization) correct what it says of their repository; anyone signed in can ask for its removal. Every request goes to OSCR's own machine, which answers it; your account page follows them.

Sign in with ORCID to claim this paper as one of its authors, correct its record or validate its tracing map: when the paper's metadata lists your ORCID iD, you are recognized at once. Maintainers of its code: sign in with GitHub, then claim the repository on your account page.

Request its removal

To ask OSCR to remove this record, the copies of its authors' scripts or its tracing map, use the removal request page: signed in, you say who you are, what to remove and why, then review and confirm the request. Published rules decide every request (how).

Discussion, reproductions, activity

Discussion: questions and error reports about this paper and its code, from signed-in readers and its authors. It opens with sign-in.

Reproductions: reports from readers who ran the authors' code: what they reproduced, with which environment, commit and data. It opens with sign-in.

Activity: what happens around this paper: new versions of its record, its map's validation, discussions and reproductions. It opens with sign-in.