OSCR

NyxBind: enhancing deep neural representations for transcription factor binding site prediction via contrastive learning.

Code ↔ Paper

14 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 14 matches · 3 of them tie a paragraph to a whole file, not to given lines: weak matches, whose lines are not tinted
  1. [1] § Materials and methods › Evaluation › Evaluation of NyxBind and alternative models for TFBS prediction ↔ evaluation/nt/nt.sh, the whole file · a weak match · score 0.79 · multi species, LoRA alpha, training epochs, NT, v2, Human
  2. [2] § Materials and methods › NyxBind overview › Contrastive learning ↔ cl/cl.py, lines 144–158 · score 0.73 · GISTEmbedLoss, MultipleNegativesRankingLoss, loss function, CL, GEL, MNRL
  3. [3] § Materials and methods › Evaluation › Evaluation of NyxBind and alternative models for TFBS prediction ↔ evaluation/cnn/cnn.py, lines 186–263 · score 0.71 · DanQ, DeepBind, hot encoding, architectures, CNN, transformer
  4. [4] § Materials and methods › Evaluation › Evaluation metrics ↔ evaluation/bert-tfbs/train.py, lines 195–216 · score 0.71 · Matthews Correlation, PR AUC, ROC AUC, recall, precision, accuracy
  5. [5] § Materials and methods › Evaluation › Evaluation metrics ↔ evaluation/cnn/cnn.py, lines 142–166 · score 0.71 · Matthews Correlation, PR AUC, ROC AUC, recall, precision, accuracy
  6. [6] § Results › NyxBind effectively identifies motifs from sequences ↔ motif/motif_benchmark/Main/GenerateMotif.py, lines 87–133 · score 0.59 · generated PWMs, motif discovery, BertSNR, JASPAR, sequences
  7. [7] § Materials and methods › NyxBind overview › Contrastive learning ↔ cl/cl.py, lines 144–158 · score 0.57 · Multiple Negatives Ranking, Symmetric, class, SMNRL, Loss, training
  8. [8] § Materials and methods › Evaluation › Fine-tuning strategies for NyxBind ↔ finetune/lora/lora.sh, the whole file · a weak match · score 0.57 · LoRA alpha, training epochs, batch
  9. [9] § Materials and methods › Evaluation › Evaluation metrics ↔ cl/cl.py, lines 169–191 · score 0.57 · Cosine AP, Sentence Transformers, metrics, model
  10. [10] § Materials and methods › Evaluation › Evaluation of NyxBind and alternative models for TFBS prediction ↔ motif/motif_benchmark/Main/TrainMultitasking.py, lines 71–135 · score 0.57 · AdamW, weight decay, optimized, loss, prediction, models
  11. [11] § Materials and methods › Evaluation › Evaluation of NyxBind and alternative models for TFBS prediction ↔ motif/motif_benchmark/Main/CrossValidToken.py, lines 1–62 · score 0.56 · AdamW, weight decay, optimized, loss, models
  12. [12] § Materials and methods › Motif analysis ↔ motif/score_from_sft.py, lines 38–63 · score 0.55 · score normalization, Attention scores, CLS, map, positions, token
  13. [13] § Materials and methods › Evaluation › Fine-tuning strategies for NyxBind ↔ finetune/ft/train.py, lines 22–29 · score 0.54 · dropout rate, LoRA alpha, FT, training
  14. [14] § Results › Training objective variants › NyxBind improves TFBS prediction ↔ evaluation/bert-tfbs/train.sh, the whole file · a weak match · score 0.50 · TFBS_N, BERT TFBS, NyxBind, DNABERT2, training, model

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 · 277 lines · 9.3 KB · no license · 3 matches

  1. import os
  2. import pandas as pd
  3. import logging
  4. import torch
  5. from datetime import datetime
  6. from datasets import Dataset
  7. from sentence_transformers import (
  8. LoggingHandler,
  9. SentenceTransformer,
  10. losses,
  11. models,
  12. SentenceTransformerTrainingArguments,
  13. SentenceTransformerTrainer
  14. )
  15. from sentence_transformers.evaluation import BinaryClassificationEvaluator
  16. from transformers import AutoModel, AutoConfig
  17. from WeightedLayerPooling import WeightedLayerPooling
  18. import argparse
  19. from torch.utils.data import DataLoader, SequentialSampler, BatchSampler
  21. #### Logging setup
  22. logging.basicConfig(
  23. format="%(asctime)s - %(message)s",
  24. datefmt="%Y-%m-%d %H:%M:%S",
  25. level=logging.INFO,
  26. handlers=[LoggingHandler()]
  27. )
  29. #### Argparse
  30. parser = argparse.ArgumentParser(description="Training script for DNABERT-2 with weighted pooling")
  31. parser.add_argument("--train_batch_size", type=int, default=128)
  32. parser.add_argument("--eval_batch_size", type=int, default=128)
  33. parser.add_argument("--num_epochs", type=int, default=3)
  34. parser.add_argument("--max_seq_length", type=int, default=30)
  35. parser.add_argument("--random_seed", type=int, default=42)
  36. parser.add_argument("--learning_rate", type=float, default=3e-5)
  37. parser.add_argument("--base_path", type=str, required=True)
  38. parser.add_argument("--model_name_or_path", type=str, required=True)
  39. parser.add_argument("--start_layer", type=int, default=11)
  40. parser.add_argument("--model_save_root", type=str, default="output/model")
  41. parser.add_argument("--loss_name", type=str, default="MNRL",
  42. choices=["MNRL", "SMNRL", "GEL", "CL"],
  43. help="Choose loss function: MNRL | SMNRL | GEL | CL")
  45. args = parser.parse_args()
  47. # ========================
  48. # Configuration
  49. # ========================
  50. train_batch_size = args.train_batch_size
  51. eval_batch_size = args.eval_batch_size
  52. num_epochs = args.num_epochs
  53. max_seq_length = args.max_seq_length
  54. random_seed = args.random_seed
  55. LR = args.learning_rate
  56. base_path = args.base_path
  57. model_name_or_path = args.model_name_or_path
  58. start_layer = args.start_layer
  59. model_save_root = args.model_save_root
  60. loss_name = args.loss_name
  62. current_time = datetime.now().strftime("%Y%m%d-%H%M%S")
  63. model_save_path = f"{model_save_root}-{loss_name}-{train_batch_size}-{num_epochs}-{LR}-{current_time}"
  64. os.makedirs(model_save_path, exist_ok=True)
  66. print(f"Base path: {base_path}")
  67. print(f"Model save path: {model_save_path}")
  68. print(f"Using loss: {loss_name}")
  70. # ========================
  71. # Dataset loader
  72. # ========================
  73. class TFCSVLoader:
  74. """Load a single CSV file with columns: sentence_A, sentence_B, label, TF"""
  75. def __new__(cls, csv_paths):
  76. self = super().__new__(cls)
  77. self.__init__(csv_paths)
  78. dataset_dict = {
  79. "sentence_A": self.data["sentence_A"].tolist(),
  80. "sentence_B": self.data["sentence_B"].tolist(),
  81. "label": self.data["label"].tolist()
  82. }
  83. dataset = Dataset.from_dict(dataset_dict)
  84. tf_dict = self.data["TF"].to_dict()
  85. return dataset, tf_dict
  87. def __init__(self, csv_paths):
  88. if len(csv_paths) != 1:
  89. raise ValueError("TFCSVLoader only supports loading one CSV file at a time.")
  90. path = csv_paths[0]
  91. if not os.path.exists(path):
  92. raise FileNotFoundError(f"{path} not found.")
  93. df = pd.read_csv(path, header=None)
  94. if df.shape[1] < 4:
  95. raise ValueError(f"{path} must contain at least 4 columns.")
  96. df.columns = ["sentence_A", "sentence_B", "label", "TF"]
  97. self.data = df
  99. # ========================
  100. # Load datasets
  101. # ========================
  102. train_csvs = [os.path.join(base_path, f) for f in os.listdir(base_path) if f.endswith("train.csv")]
  103. dev_csvs = [os.path.join(base_path, f) for f in os.listdir(base_path) if f.endswith("dev.csv")]
  104. test_csvs = [os.path.join(base_path, f) for f in os.listdir(base_path) if f.endswith("test.csv")]
  106. train_dataset, train_TF = TFCSVLoader(train_csvs)
  107. dev_dataset, dev_TF = TFCSVLoader(dev_csvs)
  108. test_dataset, test_TF = TFCSVLoader(test_csvs)
  109. print(test_dataset)
  110. print(f"Train size: {len(train_dataset)}, Dev size: {len(dev_dataset)}, Test size: {len(test_dataset)}")
  112. # ========================
  113. # Model definition
  114. # ========================
  115. class CustomTransformer(models.Transformer):
  116. """Custom transformer model to extract all hidden layers"""
  117. def __init__(self, model_name_or_path, max_seq_length, **kwargs):
  118. super().__init__(model_name_or_path, max_seq_length, **kwargs)
  119. config = AutoConfig.from_pretrained(model_name_or_path, trust_remote_code=True)
  120. self.auto_model = AutoModel.from_pretrained(model_name_or_path, config=config, trust_remote_code=True)
  122. def forward(self, features):
  123. input_ids = features["input_ids"]
  124. attention_mask = features["attention_mask"]
  125. outputs = self.auto_model(input_ids=input_ids, attention_mask=attention_mask, output_all_encoded_layers=True)
  126. hidden_states = outputs[0]
  127. features["all_layer_embeddings"] = hidden_states
  128. features["token_embeddings"] = hidden_states[-1]
  129. return features
  131. word_embedding_model = CustomTransformer(model_name_or_path, max_seq_length=max_seq_length)
  132. pooling_model = WeightedLayerPooling(
  133. word_embedding_dimension=768,
  134. num_hidden_layers=12,
  135. layer_start=start_layer
  136. )
  137. model = SentenceTransformer(modules=[word_embedding_model, pooling_model], trust_remote_code=True)
  139. # Freeze pooling layer
  140. for param in model[1].parameters():
  141. param.requires_grad = False
  142. model = model.to('cuda' if torch.cuda.is_available() else 'cpu')
  144. # ========================
  145. # Select loss function
  146. # ========================
  147. if loss_name == "MNRL":
  148. train_loss = losses.MultipleNegativesRankingLoss(model)
  149. elif loss_name == "SMNRL":
  150. train_loss = losses.MultipleNegativesSymmetricRankingLoss(model)
  151. elif loss_name == "GEL":
  152. train_loss = losses.GISTEmbedLoss(model)
  153. elif loss_name == "CL":
  154. train_loss = losses.CosineSimilarityLoss(model)
  155. else:
  156. raise ValueError(f"Unknown loss name: {loss_name}")
  158. print(f"✅ Using loss: {train_loss.__class__.__name__}")
  160. # ========================
  161. # Validation evaluator
  162. # ========================
  163. evaluator = BinaryClassificationEvaluator(
  164. sentences1=dev_dataset["sentence_A"],
  165. sentences2=dev_dataset["sentence_B"],
  166. labels=dev_dataset["label"],
  167. )
  169. # ========================
  170. # Training arguments
  171. # ========================
  172. args_sbert = SentenceTransformerTrainingArguments(
  173. output_dir=model_save_path,
  174. metric_for_best_model="cosine_ap",
  175. greater_is_better=True,
  176. num_train_epochs=num_epochs,
  177. seed=random_seed,
  178. per_device_train_batch_size=train_batch_size,
  179. per_device_eval_batch_size=eval_batch_size,
  180. learning_rate=LR,
  181. warmup_ratio=0.1,
  182. fp16=True,
  183. eval_strategy="steps",
  184. eval_steps=2000,
  185. save_strategy="steps",
  186. save_steps=2000,
  187. save_total_limit=2,
  188. load_best_model_at_end=True,
  189. logging_steps=2000,
  190. run_name=model_save_path
  191. )
  193. # ========================
  194. # Custom Trainer
  195. # ========================
  196. class MyTrainer(SentenceTransformerTrainer):
  197. """Custom Trainer using SequentialSampler for ordered batch sampling"""
  198. def train_dataloader(self):
  199. batch_sampler = BatchSampler(
  200. SequentialSampler(self.train_dataset),
  201. batch_size=self.args.per_device_train_batch_size,
  202. drop_last=False
  203. )
  204. return DataLoader(
  205. self.train_dataset,
  206. batch_sampler=batch_sampler,
  207. collate_fn=getattr(self.train_dataset, "collate_fn", None),
  208. num_workers=0
  209. )
  211. # ========================
  212. # Test evaluator
  213. # ========================
  214. test_evaluator = BinaryClassificationEvaluator(
  215. sentences1=test_dataset["sentence_A"],
  216. sentences2=test_dataset["sentence_B"],
  217. labels=test_dataset["label"],
  218. name="test",
  219. batch_size=512,
  220. )
  222. # ========================
  223. # Evaluate before training
  224. # ========================
  225. print("🔍 Evaluating on test set before training...")
  226. test_score_before = test_evaluator(model, output_path=None)
  227. print(f"✅ Test score before training: {test_score_before}")
  229. # ========================
  230. # Initialize Trainer
  231. # ========================
  232. trainer = MyTrainer(
  233. model=model,
  234. args=args_sbert,
  235. train_dataset=train_dataset,
  236. eval_dataset=dev_dataset,
  237. loss=train_loss,
  238. evaluator=evaluator
  239. )
  241. # ========================
  242. # Training
  243. # ========================
  244. trainer.train()
  245. model.save(model_save_path)
  246. print(f"✅ Model saved at {model_save_path}")
  248. # ========================
  249. # Evaluate after training
  250. # ========================
  251. print("🔍 Evaluating on test set after training...")
  252. test_score_after = test_evaluator(model, output_path=None)
  253. print(f"✅ Test score after training: {test_score_after}")
  255. # ========================
  256. # Save results
  257. # ========================
  258. eval_output_dir = "./evaloutput"
  259. os.makedirs(eval_output_dir, exist_ok=True)
  260. csv_path = os.path.join(eval_output_dir, "test_results.csv")
  262. result_row = {
  263. "timestamp": current_time,
  264. "model_path": model_save_path,
  265. "loss_name": loss_name,
  266. "test_score_before": test_score_before,
  267. "test_score_after": test_score_after
  268. }
  270. if os.path.exists(csv_path):
  271. df = pd.read_csv(csv_path)
  272. else:
  273. df = pd.DataFrame(columns=result_row.keys())
  275. df = pd.concat([df, pd.DataFrame([result_row])], ignore_index=True)
  276. df.to_csv(csv_path, index=False)
  277. print(f"✅ Test results (before & after) saved to: {csv_path}")

cl.py at commit 29d3e45, no license · at the source

Overview

Authors: Xu Yang1, Qingfa Xiao1, Yucheng Xu1, Jixin Yang1, Yusen Hou1, Weicai Long1, Miaojun Huang2, Yanlin Zhang1
ORCID iDs: Yanlin Zhang
  1. Data Science and Analytics Thrust, Information Hub, The Hong Kong University of Science and Technology (Guangzhou), No. 1 Du Xue Road, Nansha District, Guangzhou 511455, Guangdong, China
  2. College of Future Technology, The Hong Kong University of Science and Technology (Guangzhou), No. 1 Du Xue Road, Nansha District, Guangzhou 511455, Guangdong, China
Institutions: The Hong Kong University of Science and Technology (Guangzhou) (China)
Journal: Briefings in bioinformatics, volume 27, issue 2, article bbag182
Dates: received 21 October 2025; accepted 22 March 2026; published online 20 April 2026; in print March 2026
Type: Research article · Language: English
License: CC BY-NC
Identifiers: DOI 10.1093/bib/bbag182 · PMID 42007520 · PMCID PMC13093223 · OpenAlex W7154939136
Open access: gold, a free copy (OpenAlex)
Status: code verified
Categories: human (organism), cellular / molecular (subfield)
Methods: Statistics, Preprocessing, Machine learning, fMRI & imaging, Connectivity
Keywords: transcription factor binding site, genomic foundation model, contrastive learning
MeSH: Computational Biology*, Deep Learning*, Neural Networks, Computer*, Transcription Factors*, Binding Sites, Humans, Protein Binding (* major topic)
Topic: Genomics and Chromatin Dynamics (Molecular Biology, Biochemistry, Genetics and Molecular Biology), according to OpenAlex
Funding: Guangdong Provincial Project (2024QN11N085); Ministry of Human Resources and Social Security of the People's Republic of China (Y20250128)
Citations: not cited yet (Europe PMC); 36 references in the paper

Abstract

Although pretrained genomic language models effectively capture general DNA sequence patterns through masked language modeling, they often struggle to discriminate subtle yet biologically critical differences among transcription factor binding site (TFBS) motifs. Recent studies suggest that contrastive learning can enhance the discriminative power of embeddings by explicitly modeling inter-instance similarities and differences. Building on this insight, we introduce NyxBind, a TFBS prediction model that applies contrastive learning across multiple TFBS types to enhance regulatory sequence representations. Across 159 TFBS prediction tasks, NyxBind achieves the best performance on all evaluation metrics and improves Matthews Correlation Coefficient by 4.71 percentage points over DNABERT2. NyxBind supports both full-parameter and parameter-efficient fine-tuning while maintaining strong performance. It also enables accurate motif visualization, with results closely matching experimentally validated transcription factor binding profiles.

Reproduced under the paper's license (CC BY-NC), from the paper cited above.

Repository

Its files are read in the Code ↔ Paper reader above, with 14 matches between paragraphs and lines of code.

ai4nucleome/NyxBind

License: none: the authors keep all their rights
State: the link answers, verified on 30 September 2026
Evidence: files inventoried
Commit: 29d3e45f6f4961931509ad2491d9bea9cfe2c2d4, 13 March 2026
Languages: Python (39), Shell (14), Jupyter (1)
Size: 981 files, 54 scripts
Software Heritage: not archived
Found in: “Data availability”
Holds: README, environment (requirements.txt), 1 notebook
Not found: license file, CITATION.cff, tests, continuous integration, documentation
Tools: PyTorch (26 files), NumPy (24 files), Hugging Face Transformers (15 files), pandas (10 files), scikit-learn (10 files), Biopython (4 files), SciPy (3 files), Matplotlib (2 files), seaborn (1 file), statsmodels (1 file)
Availability: 1 check, the latest on 30 September 2026: the link answers
  • 30 September 2026: the link answers
55 files
At the source: github.com/ai4nucleome/NyxBind

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;
  • 54 scripts, each with its path and the digest of its content;
  • 14 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.

Data availability

The 690 ChIP-seq datasets used in this study are publicly available and can be directly downloaded from the UCSC Genome Browser ENCODE repository at https://hgdownload.cse.ucsc.edu/goldenPath/hg19/encodeDCC/wgEncodeAwgTfbsUniform/. The code for model training, fine-tuning, and motif visualization has been made publicly available at https://github.com/ai4nucleome/NyxBind. The repository also includes the datasets used for visualization.

Reproduced under the paper's license (CC BY-NC), from the paper cited above.

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, 30 September 2026: the first record

Recorded: type, language, journal, volume, issue, pages, dates, 8 authors, 3 keywords, 7 MeSH terms, 2 funders, 27 references.

Cite

This paper

Yang, X., Xiao, Q., Xu, Y., Yang, J., Hou, Y., Long, W., Huang, M., & Zhang, Y. (2026). NyxBind: enhancing deep neural representations for transcription factor binding site prediction via contrastive learning. Briefings in bioinformatics, 27(2), bbag182. https://doi.org/10.1093/bib/bbag182

BibTeX

@article{yang2026nyxbind,
author = {Yang, Xu and Xiao, Qingfa and Xu, Yucheng and Yang, Jixin and Hou, Yusen and Long, Weicai and Huang, Miaojun and Zhang, Yanlin},
title = {{NyxBind: enhancing deep neural representations for transcription factor binding site prediction via contrastive learning}},
journal = {Briefings in bioinformatics},
year = {2026},
month = mar,
volume = {27},
number = {2},
pages = {bbag182},
publisher = {Oxford University Press},
issn = {1467-5463},
doi = {10.1093/bib/bbag182},
url = {https://doi.org/10.1093/bib/bbag182},
pmid = {42007520},
pmcid = {PMC13093223}
}

RIS

TY - JOUR
AU - Yang, Xu
AU - Xiao, Qingfa
AU - Xu, Yucheng
AU - Yang, Jixin
AU - Hou, Yusen
AU - Long, Weicai
AU - Huang, Miaojun
AU - Zhang, Yanlin
TI - NyxBind: enhancing deep neural representations for transcription factor binding site prediction via contrastive learning
T2 - Briefings in bioinformatics
J2 - Brief Bioinform
PY - 2026
DA - 2026/03/01
VL - 27
IS - 2
SP - bbag182
SN - 1467-5463
PB - Oxford University Press
DO - 10.1093/bib/bbag182
UR - https://doi.org/10.1093/bib/bbag182
LA - en
ER -

CSL-JSON

{
"id": "10.1093/bib/bbag182",
"type": "article-journal",
"title": "NyxBind: enhancing deep neural representations for transcription factor binding site prediction via contrastive learning",
"container-title": "Briefings in bioinformatics",
"author": [
{
"family": "Yang",
"given": "Xu"
},
{
"family": "Xiao",
"given": "Qingfa"
},
{
"family": "Xu",
"given": "Yucheng"
},
{
"family": "Yang",
"given": "Jixin"
},
{
"family": "Hou",
"given": "Yusen"
},
{
"family": "Long",
"given": "Weicai"
},
{
"family": "Huang",
"given": "Miaojun"
},
{
"family": "Zhang",
"given": "Yanlin"
}
],
"container-title-short": "Brief Bioinform",
"volume": "27",
"issue": "2",
"page": "bbag182",
"DOI": "10.1093/bib/bbag182",
"PMID": "42007520",
"PMCID": "PMC13093223",
"ISSN": "1467-5463",
"publisher": "Oxford University Press",
"URL": "https://doi.org/10.1093/bib/bbag182",
"language": "en",
"issued": {
"date-parts": [
[
2026,
3,
1
]
]
}
}

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:10.1038/s41592-026-03057-2 [code]
CREsted: modeling genomic and synthetic cell-type-specific enhancers across tissues and species.
Journal: Nature methods
In common: Biopython, statsmodels, PyTorch, 6 other tools, 5 references
[2] doi:10.1093/nargab/lqag050 [code]
TSProm: deep learning framework to predict tissue-specific regulatory logic.
Journal: NAR genomics and bioinformatics
In common: Biopython, Hugging Face Transformers, statsmodels, 7 other tools, 4 references
[3] doi:10.1038/s42003-026-10462-y [code]
SpaDC enables sequence-based integrative analysis and regulatory inference of spatial chromatin accessibility data.
Journal: Communications biology
In common: Biopython, PyTorch, scikit-learn, 4 other tools, 3 references
[4] doi:10.1038/s42003-026-10957-8 [code]
Brain defence by the extracellular matrix protein Cochlin.
Journal: Communications biology
In common: Biopython, Hugging Face Transformers, statsmodels, 7 other tools, cellular / molecular
[5] doi:10.64898/2026.03.30.714220 [code]
An integrated single cell and spatial omics atlas of human prenatal development
Journal: bioRxiv (preprint)
In common: Biopython, Hugging Face Transformers, statsmodels, 7 other tools
[6] doi:10.1093/bib/bbag339 [code]
scDeepAPA: a deep learning framework for single-cell alternative polyadenylation identification.
Journal: Briefings in bioinformatics
In common: Biopython, Hugging Face Transformers, PyTorch, 5 other tools, 1 reference
[7] doi:10.1021/acsomega.5c09368 [code]
Structure-Based and AI-Assisted Identification of AGPS Inhibitors for Glioma via Integrated Docking, Molecular Dynamics, and Binding Affinity Screening.
Journal: ACS omega
In common: Biopython, Hugging Face Transformers, PyTorch, 6 other tools
[8] doi:10.1186/s12864-026-12965-8 [code]
Systematic evaluation of single-cell foundation model interpretability: attention-derived edge scores add no incremental value over gene-level features for perturbation-target prediction.
Journal: BMC genomics
In common: Hugging Face Transformers, statsmodels, PyTorch, 6 other tools, cellular / molecular
[9] doi:10.64898/2026.03.02.709014 [code]
A Protein Language Model Reveals Organellar Ca <sup>2+</sup> ATPases at Neuronal Synapses
Journal: bioRxiv (preprint)
In common: Hugging Face Transformers, statsmodels, PyTorch, 6 other tools, cellular / molecular
[10] doi:10.1016/j.isci.2026.117180 [code]
Developmental changes in similarity between neural representations of mental arithmetic and artificial neural networks.
Journal: iScience
In common: Hugging Face Transformers, statsmodels, PyTorch, 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.