OSCR

DAG-VAERL: a novel causal inference method for building causal gene regulatory networks.

Code ↔ Paper

5 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 5 matches
  1. [1] § Experiments › Synthetic datasets ↔ utils.py, lines 53–98 · score 0.74 · linear exp, linear gumbel, linear gauss, Exponential, Noise, sem
  2. [2] § Method › Reinforcement learning formulation and policy network › Training and optimization process ↔ rl_training.py, lines 14–95 · score 0.56 · Smooth L1 loss, policy, training, probabilities, optimized
  3. [3] § Experiments ↔ visualization.py, lines 128–153 · score 0.55 · evaluation metrics, F1 score, Recall, Precision, DAG
  4. [4] § Experiments › Synthetic datasets ↔ utils.py, lines 53–98 · score 0.55 · noise_scale, x_dims, simulating, sem
  5. [5] § Method › Reinforcement learning formulation and policy network ↔ network.py, lines 180–237 · score 0.53 · RLDAGPolicy, edge probabilities, module, weights, network

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 · 493 lines · 16 KB · no license · 2 matches

  1. import numpy as np
  2. import torch
  3. from torch.utils.data.dataset import TensorDataset
  4. from torch.utils.data import DataLoader
  5. import torch.nn.functional as F
  6. import torch.nn as nn
  7. from torch.autograd import Variable
  8. import scipy.linalg as slin
  9. import scipy.sparse as sp
  10. import networkx as nx
  11. import pandas as pd
  12. import os
  13. import glob
  14. import re
  15. import pickle
  16. import math
  17. from torch.optim.adam import Adam
  18. ---------
  20. def simulate_random_dag(d: int,
  21. degree: float,
  22. graph_type: str,
  23. w_range: tuple = (0.5, 2.0)) -> nx.DiGraph:
  25. if graph_type == 'erdos-renyi':
  26. prob = float(degree) / (d - 1)
  27. B = np.tril((np.random.rand(d, d) < prob).astype(float), k=-1)
  28. elif graph_type == 'barabasi-albert':
  29. m = int(round(degree / 2))
  30. B = np.zeros([d, d])
  31. bag = [0]
  32. for ii in range(1, d):
  33. dest = np.random.choice(bag, size=m)
  34. for jj in dest:
  35. B[ii, jj] = 1
  36. bag.append(ii)
  37. bag.extend(dest)
  38. elif graph_type == 'full':
  40. B = np.tril(np.ones([d, d]), k=-1)
  41. else:
  42. raise ValueError('unknown graph type')
  44. P = np.random.permutation(np.eye(d, d))
  45. B_perm = P.T.dot(B).dot(P)
  46. U = np.random.uniform(low=w_range[0], high=w_range[1], size=[d, d])
  47. U[np.random.rand(d, d) < 0.5] *= -1
  48. W = (B_perm != 0).astype(float) * U
  49. G = nx.DiGraph(W)
  50. return G
  53. def simulate_sem(G: nx.DiGraph,
  54. n: int, x_dims: int,
  55. sem_type: str,
  56. linear_type: str,
  57. noise_scale: float = 1.0) -> np.ndarray:
  58. W = nx.to_numpy_array(G)
  59. d = W.shape[0]
  60. X = np.zeros([n, d, x_dims])
  61. ordered_vertices = list(nx.topological_sort(G))
  62. assert len(ordered_vertices) == d
  64. for j in ordered_vertices:
  65. parents = list(G.predecessors(j))
  66. if linear_type == 'linear':
  67. eta = X[:, parents, 0].dot(W[parents, j])
  68. elif linear_type == 'nonlinear_1':
  69. eta = np.cos(X[:, parents, 0] + 1).dot(W[parents, j])
  70. elif linear_type == 'nonlinear_2':
  71. eta = (X[:, parents, 0] + 0.5).dot(W[parents, j])
  72. else:
  73. raise ValueError('unknown linear data type')
  75. if sem_type == 'linear-gauss':
  76. if linear_type == 'linear':
  77. X[:, j, 0] = eta + np.random.normal(scale=noise_scale, size=n)
  78. elif linear_type == 'nonlinear_1':
  79. X[:, j, 0] = eta + np.random.normal(scale=noise_scale, size=n)
  80. elif linear_type == 'nonlinear_2':
  81. X[:, j, 0] = 2. * np.sin(eta) + eta + np.random.normal(scale=noise_scale, size=n)
  82. elif sem_type == 'linear-exp':
  83. X[:, j, 0] = eta + np.random.exponential(scale=noise_scale, size=n)
  84. elif sem_type == 'linear-gumbel':
  85. X[:, j, 0] = eta + np.random.gumbel(scale=noise_scale, size=n)
  86. else:
  87. raise ValueError('unknown sem type')
  90. if x_dims > 1:
  91. for i in range(x_dims - 1):
  92. X[:, :, i+1] = (np.random.normal(scale=noise_scale, size=1) * X[:, :, 0]
  93. + np.random.normal(scale=noise_scale, size=1)
  94. + np.random.normal(scale=noise_scale, size=(n, d)))
  95. X[:, :, 0] = (np.random.normal(scale=noise_scale, size=1) * X[:, :, 0]
  96. + np.random.normal(scale=noise_scale, size=1)
  97. + np.random.normal(scale=noise_scale, size=(n, d)))
  98. return X
  101. def simulate_population_sample(W: np.ndarray,
  102. Omega: np.ndarray) -> np.ndarray:
  104. d = W.shape[0]
  105. X = np.sqrt(d) * slin.sqrtm(Omega).dot(np.linalg.pinv(np.eye(d) - W))
  106. return X
  109. def count_accuracy(G_true: nx.DiGraph,
  110. G: nx.DiGraph,
  111. G_und: nx.DiGraph = None) -> tuple:
  113. B_true = nx.to_numpy_array(G_true) != 0
  114. B = nx.to_numpy_array(G) != 0
  115. B_und = None if G_und is None else nx.to_numpy_array(G_und)
  116. d = B.shape[0]
  118. if B_und is not None:
  119. pred_und = np.flatnonzero(B_und)
  120. pred = np.flatnonzero(B)
  121. cond = np.flatnonzero(B_true)
  122. cond_reversed = np.flatnonzero(B_true.T)
  123. cond_skeleton = np.concatenate([cond, cond_reversed])
  125. true_pos = np.intersect1d(pred, cond, assume_unique=True)
  126. if B_und is not None:
  127. true_pos_und = np.intersect1d(pred_und, cond_skeleton, assume_unique=True)
  128. true_pos = np.concatenate([true_pos, true_pos_und])
  130. false_pos = np.setdiff1d(pred, cond_skeleton, assume_unique=True)
  131. if B_und is not None:
  132. false_pos_und = np.setdiff1d(pred_und, cond_skeleton, assume_unique=True)
  133. false_pos = np.concatenate([false_pos, false_pos_und])
  136. extra = np.setdiff1d(pred, cond, assume_unique=True)
  137. reverse = np.intersect1d(extra, cond_reversed, assume_unique=True)
  139. pred_size = len(pred)
  140. if B_und is not None:
  141. pred_size += len(pred_und)
  142. cond_neg_size = 0.5 * d * (d - 1) - len(cond)
  144. fdr = float(len(reverse) + len(false_pos)) / max(pred_size, 1)
  145. tpr = float(len(true_pos)) / max(len(cond), 1)
  146. fpr = float(len(reverse) + len(false_pos)) / max(cond_neg_size, 1)
  149. B_lower = np.tril(B + B.T)
  150. if B_und is not None:
  151. B_lower += np.tril(B_und + B_und.T)
  152. pred_lower = np.flatnonzero(B_lower)
  153. cond_lower = np.flatnonzero(np.tril(B_true + B_true.T))
  154. extra_lower = np.setdiff1d(pred_lower, cond_lower, assume_unique=True)
  155. missing_lower = np.setdiff1d(cond_lower, pred_lower, assume_unique=True)
  156. shd = len(extra_lower) + len(missing_lower) + len(reverse)
  158. return fdr, tpr, fpr, shd, pred_size
  161. def read_BNrep(args):
  162. """load results from BN repository"""
  163. if args.data_filename == 'alarm':
  164. data_dir = os.path.join(args.data_dir, 'alarm/')
  165. elif args.data_filename == 'child':
  166. data_dir = os.path.join(args.data_dir, 'child/')
  167. elif args.data_filename == 'hail':
  168. data_dir = os.path.join(args.data_dir, 'hail/')
  169. elif args.data_filename == 'alarm10':
  170. data_dir = os.path.join(args.data_dir, 'alarm10/')
  171. elif args.data_filename == 'child10':
  172. data_dir = os.path.join(args.data_dir, 'child10/')
  173. elif args.data_filename == 'pigs':
  174. data_dir = os.path.join(args.data_dir, 'pigs/')
  175. else:
  176. raise ValueError("Unknown data_filename for BN repository")
  178. all_data = dict()
  179. file_pattern = data_dir + "*_s*_v*.txt"
  180. all_files = glob.iglob(file_pattern)
  182. for file in all_files:
  183. match = re.search(r'/([\w]+)_s([\w]+)_v([\w]+).txt', file)
  184. dataset, samplesN, version = match.group(1), match.group(2), match.group(3)
  185. data = np.loadtxt(file, skiprows=0, dtype=np.int32)
  186. if samplesN not in all_data:
  187. all_data[samplesN] = dict()
  188. all_data[samplesN][version] = data
  191. file_pattern = data_dir + "*_graph.txt"
  192. files = glob.iglob(file_pattern)
  193. graph = None
  194. for f in files:
  195. graph_data = np.loadtxt(f, skiprows=0, dtype=np.int32)
  196. graph = graph_data
  198. return all_data, graph
  201. def load_data(args, batch_size=1000, suffix='', debug=False):
  202. n, d = args.data_sample_size, args.data_variable_size
  203. graph_type = args.graph_type
  204. degree = args.graph_degree
  205. sem_type = args.graph_sem_type
  206. linear_type = args.graph_linear_type
  207. x_dims = args.x_dims
  209. if args.data_type == 'synthetic':
  210. # generate data
  211. G = simulate_random_dag(d, degree, graph_type)
  212. X = simulate_sem(G, n, x_dims, sem_type, linear_type)
  214. elif args.data_type == 'discrete':
  215. if args.data_filename.endswith('.pkl'):
  216. with open(os.path.join(args.data_dir, args.data_filename), 'rb') as handle:
  217. X = pickle.load(handle)
  218. G = None
  219. else:
  220. all_data, graph = read_BNrep(args)
  221. G = nx.DiGraph(graph)
  222. X = all_data['1000']['1']
  223. else:
  224. raise ValueError("Unknown data_type, must be 'synthetic' or 'discrete'")
  226. feat_train = torch.FloatTensor(X)
  227. feat_valid = torch.FloatTensor(X)
  228. feat_test = torch.FloatTensor(X)
  230. train_data = TensorDataset(feat_train, feat_train)
  231. valid_data = TensorDataset(feat_valid, feat_train)
  232. test_data = TensorDataset(feat_test, feat_train)
  234. train_data_loader = DataLoader(train_data, batch_size=batch_size)
  235. valid_data_loader = DataLoader(valid_data, batch_size=batch_size)
  236. test_data_loader = DataLoader(test_data, batch_size=batch_size)
  238. return train_data_loader, valid_data_loader, test_data_loader, G
  241. def encode_onehot(labels):
  242. classes = set(labels)
  243. classes_dict = {c: np.identity(len(classes))[i, :] for i, c in enumerate(classes)}
  244. labels_onehot = np.array(list(map(classes_dict.get, labels)), dtype=np.int32)
  245. return labels_onehot
  248. def my_softmax(input, axis=1):
  249. trans_input = input.transpose(axis, 0).contiguous()
  250. soft_max_1d = F.softmax(trans_input, dim=0)
  251. return soft_max_1d.transpose(axis, 0)
  254. def binary_concrete(logits, tau=1, hard=False, eps=1e-10):
  255. y_soft = binary_concrete_sample(logits, tau=tau, eps=eps)
  256. if hard:
  257. y_hard = (y_soft > 0.5).float()
  258. y = Variable(y_hard.data - y_soft.data) + y_soft
  259. else:
  260. y = y_soft
  261. return y
  264. def binary_concrete_sample(logits, tau=1, eps=1e-10):
  265. logistic_noise = sample_logistic(logits.size(), eps=eps)
  266. if logits.is_cuda:
  267. logistic_noise = logistic_noise.cuda()
  268. y = logits + Variable(logistic_noise)
  269. return torch.sigmoid(y / tau)
  272. def sample_logistic(shape, eps=1e-10):
  273. uniform = torch.rand(shape).float()
  274. return torch.log(uniform + eps) - torch.log(1 - uniform + eps)
  277. def sample_gumbel(shape, eps=1e-10):
  278. U = torch.rand(shape).float()
  279. return - torch.log(eps - torch.log(U + eps))
  282. def gumbel_softmax_sample(logits, tau=1, eps=1e-10):
  283. gumbel_noise = sample_gumbel(logits.size(), eps=eps).double()
  284. if logits.is_cuda:
  285. gumbel_noise = gumbel_noise.cuda()
  286. y = logits + Variable(gumbel_noise)
  287. return my_softmax(y / tau, axis=-1)
  290. def gumbel_softmax(logits, tau=1, hard=False, eps=1e-10):
  291. y_soft = gumbel_softmax_sample(logits, tau=tau, eps=eps)
  292. if hard:
  293. shape = logits.size()
  294. _, k = y_soft.data.max(-1)
  295. y_hard = torch.zeros(*shape, dtype=torch.double, device=logits.device)
  296. y_hard.scatter_(-1, k.view(shape[:-1] + (1,)), 1.0)
  297. y = Variable(y_hard - y_soft.data) + y_soft
  298. else:
  299. y = y_soft
  300. return y
  303. def gauss_sample_z(logits, zsize):
  304. U = torch.randn(logits.size(0), zsize).double()
  305. x = torch.zeros(logits.size(0), 1, zsize).double()
  306. for j in range(logits.size(0)):
  307. x[j, 0, :] = U[j, :] * torch.exp(logits[j, 0, zsize:2*zsize]) + logits[j, 0, 0:zsize]
  308. return x
  311. def gauss_sample_z_new(logits, zsize):
  312. U = torch.randn(logits.size(0), logits.size(1), zsize).double()
  313. x = torch.zeros(logits.size(0), logits.size(1), zsize).double()
  314. x[:, :, :] = U[:, :, :] + logits[:, :, 0:zsize]
  315. return x
  318. def binary_accuracy(output, labels):
  319. preds = output > 0.5
  320. correct = preds.type_as(labels).eq(labels).double()
  321. correct = correct.sum()
  322. return correct / len(labels)
  325. def kl_categorical(preds, log_prior, num_atoms, eps=1e-16):
  326. kl_div = preds * (torch.log(preds + eps) - torch.log(log_prior + eps))
  327. return kl_div.sum() / (num_atoms)
  330. def kl_gaussian(preds, zsize):
  331. predsnew = preds.squeeze(1)
  332. mu = predsnew[:, 0:zsize]
  333. log_sigma = predsnew[:, zsize:2*zsize]
  334. kl_div = torch.exp(2*log_sigma) - 2*log_sigma + mu * mu
  335. kl_sum = kl_div.sum()
  336. return (kl_sum / (preds.size(0)) - zsize) * 0.5
  339. def kl_gaussian_sem(preds):
  340. mu = preds
  341. kl_div = mu * mu
  342. kl_sum = kl_div.sum()
  343. return (kl_sum / (preds.size(0))) * 0.5
  346. def kl_categorical_uniform(preds, num_atoms, num_edge_types, add_const=False, eps=1e-16):
  347. kl_div = preds * torch.log(preds + eps)
  348. if add_const:
  349. const = np.log(num_edge_types)
  350. kl_div += const
  351. return kl_div.sum() / (num_atoms * preds.size(0))
  354. def nll_gaussian(preds, target, variance, add_const=False):
  356. if isinstance(variance, float):
  357. variance = torch.tensor(variance, dtype=preds.dtype, device=preds.device)
  359. neg_log_p = variance + (preds - target)**2 / (2. * torch.exp(2. * variance))
  361. if add_const:
  362. const = 0.5 * torch.log(2 * torch.from_numpy(np.pi).to(variance.device) * variance)
  363. neg_log_p += const
  365. return neg_log_p.sum() / (target.size(0))
  370. def normalize_adj(adj):
  371. rowsum = torch.abs(torch.sum(adj, 1))
  372. d_inv_sqrt = torch.pow(rowsum, -0.5)
  373. d_inv_sqrt[torch.isinf(d_inv_sqrt)] = 0.
  374. d_mat_inv_sqrt = torch.diag(d_inv_sqrt)
  375. myr = torch.matmul(torch.matmul(d_mat_inv_sqrt, adj), d_mat_inv_sqrt)
  376. myr[isnan(myr)] = 0.
  377. return myr
  380. def preprocess_adj(adj):
  381. device = adj.device
  382. I = torch.eye(adj.shape[0], dtype=torch.double, device=device)
  383. adj_normalized = I + adj.transpose(0, 1)
  384. return adj_normalized
  387. def preprocess_adj_new(adj):
  388. device = adj.device
  389. I = torch.eye(adj.shape[0], dtype=torch.double, device=device)
  390. adj_normalized = I - adj.transpose(0, 1)
  391. return adj_normalized
  394. def preprocess_adj_new1(adj):
  395. device = adj.device
  396. I = torch.eye(adj.shape[0], dtype=torch.double, device=device)
  397. adj_normalized = torch.inverse(I - adj.transpose(0, 1))
  398. return adj_normalized
  401. def isnan(x):
  402. return x != x
  405. def my_normalize(z):
  406. device = z.device
  407. znor = torch.zeros(z.size(), dtype=torch.double, device=device)
  408. for i in range(z.size(0)):
  409. testnorm = torch.norm(z[i, :, :], dim=0)
  410. znor[i, :, :] = z[i, :, :] / testnorm
  411. znor[isnan(znor)] = 0.0
  412. return znor
  415. def sparse_to_tuple(sparse_mx):
  416. def to_tuple(mx):
  417. if not sp.isspmatrix_coo(mx):
  418. mx = mx.tocoo()
  419. coords = np.vstack((mx.row, mx.col)).transpose()
  420. values = mx.data
  421. shape = mx.shape
  422. return coords, values, shape
  424. if isinstance(sparse_mx, list):
  425. for i in range(len(sparse_mx)):
  426. sparse_mx[i] = to_tuple(sparse_mx[i])
  427. else:
  428. sparse_mx = to_tuple(sparse_mx)
  430. return sparse_mx
  433. def matrix_poly(matrix, d):
  434. device = matrix.device
  435. I = torch.eye(d, dtype=torch.double, device=device)
  436. x = I + matrix / d
  437. return torch.matrix_power(x, d)
  440. def A_connect_loss(A, tol, z):
  441. d = A.size()[0]
  442. loss = 0
  443. for i in range(d):
  444. loss += 2 * tol - torch.sum(torch.abs(A[:, i])) - torch.sum(torch.abs(A[i, :])) + z * z
  445. return loss
  448. def A_positive_loss(A, z_positive):
  449. result = -A + z_positive * z_positive
  450. loss = torch.sum(result)
  451. return loss
  455. def compute_BiCScore(G, D):
  456. origin_score = []
  457. num_var = G.shape[0]
  458. for i in range(num_var):
  459. parents = np.where(G[:, i] != 0)
  460. score_one = compute_local_BiCScore(D, i, parents)
  461. origin_score.append(score_one)
  462. score = sum(origin_score)
  463. return score
  466. def compute_local_BiCScore(np_data, target, parents):
  467. sample_size = np_data.shape[0]
  468. count_d = dict()
  469. for data_ind in range(sample_size):
  470. parent_combination = tuple(np_data[data_ind, parents].reshape(1, -1)[0])
  471. self_value = tuple(np_data[data_ind, target].reshape(1, -1)[0])
  472. if parent_combination in count_d:
  473. if self_value in count_d[parent_combination]:
  474. count_d[parent_combination][self_value] += 1.0
  475. else:
  476. count_d[parent_combination][self_value] = 1.0
  477. else:
  478. count_d[parent_combination] = dict()
  479. count_d[parent_combination][self_value] = 1.0
  481. loglik = 0.0
  482. num_parent_state = np.prod(np.amax(np_data[:, parents], axis=0) + 1)
  483. num_self_state = np.amax(np_data[:, target], axis=0) + 1
  485. for parents_state in count_d:
  486. local_count = sum(count_d[parents_state].values())
  487. for self_state in count_d[parents_state]:
  488. loglik += count_d[parents_state][self_state] * (
  489. math.log(count_d[parents_state][self_state] + 0.1) - math.log(local_count))
  491. num_param = num_parent_state * (num_self_state - 1)
  492. bic = loglik - 0.5 * math.log(sample_size) * num_param
  493. return bic

utils.py at commit 533ee14, no license · at the source

Overview

Authors: Teng Long1, Sachit Satyal1, Yong-Fang Kuo2, Jean Gao3
  1. Department of Computer Science and Engineering, University of Texas at Arlington,500 UTA Blvd., Arlington, TX 76019 USA
  2. Department of Biostatistics and Data Science, University of Texas Medical Branch,301 University Blvd., Galveston, TX 77555 USA
  3. Department of Computer Science, Baylor University,One Bear Place #97141, Waco, TX 76798 USA
Institutions: The University of Texas at Arlington (United States); The University of Texas Medical Branch at Galveston (United States); Baylor University (United States)
Journal: BioData mining, volume 19, issue 1, article 68
Dates: received 1 December 2025; accepted 27 May 2026; published online 16 June 2026
Type: Research article · Language: English
License: CC BY-NC-ND
Identifiers: DOI 10.1186/s13040-026-00571-z · PMID 42304400 · PMCID PMC13520353 · OpenAlex W7164936735
Open access: gold, a free copy (OpenAlex)
Status: code verified
Categories: Alzheimer's / dementia (population)
Methods: Machine learning
Keywords: Causality, Directed acyclic graph, Deep learning, Reinforcement learning, Long non-coding RNAs, Alzheimer’s disease
Topic: Bayesian Modeling and Causal Inference (Artificial Intelligence, Computer Science), according to OpenAlex
Citations: not cited yet (Europe PMC); 28 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 5 matches between paragraphs and lines of code.

tenglong322/DAG-VAERL

License: none: the authors keep all their rights
State: the link answers, verified on 27 September 2026
Evidence: files inventoried
Commit: 533ee14130b88ec3f707e7cb379c0e5653b5d8c4, 6 May 2026
Languages: Python (10)
Size: 11 files, 10 scripts
Software Heritage: not archived
Found in: “Data availability”
Holds: README
Not found: license file, CITATION.cff, environment file, tests, continuous integration, documentation
Tools: NumPy (9 files), PyTorch (8 files), NetworkX (6 files), pandas (4 files), Matplotlib (2 files), SciPy (1 file)
Availability: 1 check, the latest on 27 September 2026: the link answers
  • 27 September 2026: the link answers
11 files
At the source: github.com/tenglong322/DAG-VAERL

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;
  • 10 scripts, each with its path and the digest of its content;
  • 5 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:

Read it in the paper: doi.org/10.1186/s13040-026-00571-z.

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, 27 September 2026: the first record

Recorded: type, language, journal, volume, issue, pages, dates, 4 authors, 6 keywords, 15 references.

Cite

This paper

Long, T., Satyal, S., Kuo, Y.-F., & Gao, J. (2026). DAG-VAERL: a novel causal inference method for building causal gene regulatory networks. BioData mining, 19(1), 68. https://doi.org/10.1186/s13040-026-00571-z

BibTeX

@article{long2026dag,
author = {Long, Teng and Satyal, Sachit and Kuo, Yong-Fang and Gao, Jean},
title = {{DAG-VAERL: a novel causal inference method for building causal gene regulatory networks}},
journal = {BioData mining},
year = {2026},
month = jun,
volume = {19},
number = {1},
pages = {68},
publisher = {BMC},
issn = {1756-0381},
doi = {10.1186/s13040-026-00571-z},
url = {https://doi.org/10.1186/s13040-026-00571-z},
pmid = {42304400},
pmcid = {PMC13520353}
}

RIS

TY - JOUR
AU - Long, Teng
AU - Satyal, Sachit
AU - Kuo, Yong-Fang
AU - Gao, Jean
TI - DAG-VAERL: a novel causal inference method for building causal gene regulatory networks
T2 - BioData mining
J2 - BioData Min
PY - 2026
DA - 2026/06/16
VL - 19
IS - 1
SP - 68
SN - 1756-0381
PB - BMC
DO - 10.1186/s13040-026-00571-z
UR - https://doi.org/10.1186/s13040-026-00571-z
LA - en
ER -

CSL-JSON

{
"id": "10.1186/s13040-026-00571-z",
"type": "article-journal",
"title": "DAG-VAERL: a novel causal inference method for building causal gene regulatory networks",
"container-title": "BioData mining",
"author": [
{
"family": "Long",
"given": "Teng"
},
{
"family": "Satyal",
"given": "Sachit"
},
{
"family": "Kuo",
"given": "Yong-Fang"
},
{
"family": "Gao",
"given": "Jean"
}
],
"container-title-short": "BioData Min",
"volume": "19",
"issue": "1",
"page": "68",
"DOI": "10.1186/s13040-026-00571-z",
"PMID": "42304400",
"PMCID": "PMC13520353",
"ISSN": "1756-0381",
"publisher": "BMC",
"URL": "https://doi.org/10.1186/s13040-026-00571-z",
"language": "en",
"issued": {
"date-parts": [
[
2026,
6,
16
]
]
}
}

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/s41514-026-00443-0 [code]
Network-based discovery of regulatory drivers of cognitive decline in alzheimer's disease.
Journal: npj aging
In common: NetworkX, pandas, SciPy, 2 other tools, Alzheimer's / dementia, 1 reference
[2] doi:10.1093/bioinformatics/btag553 [code]
Spatial-spectral fusion enables drug repositioning by capturing indirect and long-range associations in biological networks.
Journal: Bioinformatics (Oxford, England)
In common: NetworkX, PyTorch, pandas, 3 other tools, Alzheimer's / dementia
[3] doi:10.1002/dad2.70430 [code]
Benchmarking privacy and utility in synthetic tabular cohorts for Alzheimer's disease research.
Journal: Alzheimer's & dementia (Amsterdam, Netherlands)
In common: NetworkX, PyTorch, pandas, 3 other tools, Alzheimer's / dementia
[4] doi:10.21203/rs.3.rs-9060414/v1 [code]
Cross-Species Aging Knowledge Integration into Agentic AI Platform Uncovers Conserved Mechanisms
Journal: Research Square (preprint)
In common: NetworkX, PyTorch, pandas, 3 other tools, Alzheimer's / dementia
[5] doi:10.1038/s44400-026-00094-8 [code]
Haplotype-resolved DNA methylation at the &lt;i&gt;APOE&lt;/i&gt; locus identifies allele-specific epigenetic signatures relevant to Alzheimer's disease risk.
Journal: NPJ dementia
In common: PyTorch, pandas, SciPy, 2 other tools, Alzheimer's / dementia, 1 reference
[6] doi:10.1038/s41467-026-71267-5 [code]
Human-like cognitive generalization for large models via mental representation-guided supervision.
Journal: Nature communications
In common: PyTorch, pandas, SciPy, 2 other tools, 1 reference
[7] doi:10.7554/elife.110588 [code]
Opening the black box toward a modular approach to spike sorting.
Journal: eLife
In common: NetworkX, PyTorch, pandas, 3 other tools
[8] doi:10.1038/s42003-026-10957-8 [code]
Brain defence by the extracellular matrix protein Cochlin.
Journal: Communications biology
In common: NetworkX, PyTorch, pandas, 3 other tools
[9] doi:10.7554/elife.108950 [code]
Comprehensive RNA velocity by modeling the cascade of gene regulation, transcription, and splicing from single-cell RNA sequencing data with TSvelo.
Journal: eLife
In common: NetworkX, PyTorch, pandas, 3 other tools
[10] doi:10.1126/sciadv.aee9425 [code]
Probabilistic inference of homonymous and heteronymous recurrent inhibition in human muscles from large-scale motor neuron recordings.
Journal: Science advances
In common: NetworkX, PyTorch, pandas, 3 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.