THGB: predicting ligand-receptor interactions by combining tree boosting and histogram-based gradient boosting DOI Creative Commons
Liqian Zhou, Jiao Song, Zejun Li

и другие.

Scientific Reports, Год журнала: 2024, Номер 14(1)

Опубликована: Ноя. 28, 2024

Ligand-receptor interaction (LRI) prediction has great significance in biological and medical research facilitates to infer analyze cell-to-cell communication. However, wet experiments for new LRI discovery are costly time-consuming. Here, we propose a computational model called THGB uncover LRIs. first extracts feature information of Ligand-Receptor (LR) pairs using iFeature. Next, it adopts tree boosting obtain representative LR features. Finally, devises the histogram-based gradient capture high-quality To assess performance, compared with three models (i.e., CellEnBoost, CellGiQ, CellComNet) one classical protein-protein inference PIPR. The results demonstrated that achieved best overall predictions terms six evaluation indictors precision, recall, accuracy, F1-score, AUC, AUPR). measure effect selection on prediction, was four methods PCA, NMF, LLE, TSVD). showed more appropriate select features improve prediction. We also conducted ablation study found outperformed without selection. hope is useful tool find LRIs further

Язык: Английский

Identifying potential ligand–receptor interactions based on gradient boosted neural network and interpretable boosting machine for intercellular communication analysis DOI
Lihong Peng, Pengfei Gao, Wei Xiong

и другие.

Computers in Biology and Medicine, Год журнала: 2024, Номер 171, С. 108110 - 108110

Опубликована: Фев. 6, 2024

Язык: Английский

Процитировано

12
MGNDTI: A Drug-Target Interaction Prediction Framework Based on Multimodal Representation Learning and the Gating Mechanism DOI
Lihong Peng, Xin Liu, Min Chen

и другие.

Journal of Chemical Information and Modeling, Год журнала: 2024, Номер 64(16), С. 6684 - 6698

Опубликована: Авг. 13, 2024

Drug-Target Interaction (DTI) prediction facilitates acceleration of drug discovery and promotes repositioning. Most existing deep learning-based DTI methods can better extract discriminative features for drugs proteins, but they rarely consider multimodal drugs. Moreover, learning the interaction representations between targets needs further exploration. Here, we proposed a simple M ulti-modal G ating N etwork prediction, MGNDTI, based on representation gating mechanism. MGNDTI first learns sequence using different retentive networks. Next, it extracts molecular graph through convolutional network. Subsequently, devises network to obtain joint targets. Finally, builds fully connected computing probability. was benchmarked against seven state-of-the-art models (CPI-GNN, TransformerCPI, MolTrans, BACPI, CPGL, GIFDTI, FOTF-CPI) four data sets (i.e., Human, C. elegans, BioSNAP, BindingDB) under experimental settings. Through evaluation with AUROC, AUPRC, accuracy, F1 score, MCC, significantly outperformed above methods. is powerful tool showcasing its superior robustness generalization ability diverse It freely available at https://github.com/plhhnu/MGNDTI.

Язык: Английский

Процитировано

6
Unveiling patterns in spatial transcriptomics data: a novel approach utilizing graph attention autoencoder and multiscale deep subspace clustering network DOI Creative Commons
Liqian Zhou,

Xinhuai Peng,

Min Chen

и другие.

GigaScience, Год журнала: 2025, Номер 14

Опубликована: Янв. 1, 2025

Abstract Background The accurate deciphering of spatial domains, along with the identification differentially expressed genes and inference cellular trajectory based on transcriptomic (ST) data, holds significant potential for enhancing our understanding tissue organization biological functions. However, most clustering methods can neither decipher complex structures in ST data nor entirely employ features embedded different layers. Results This article introduces STMSGAL, a novel framework analyzing by incorporating graph attention autoencoder multiscale deep subspace clustering. First, STMSGAL constructs ctaSNN, cell type–aware shared nearest neighbor graph, using Louvian exclusively gene expression profiles. Subsequently, it integrates profiles ctaSNN to generate spot latent representations Lastly, implements clustering, differential analysis, inference, providing comprehensive capabilities thorough exploration interpretation. was evaluated against 7 methods, including SCANPY, SEDR, CCST, DeepST, GraphST, STAGATE, SiGra, four 10x Genomics Visium datasets, 1 mouse visual cortex STARmap dataset, 2 Stereo-seq embryo datasets. comparison showcased STMSGAL’s remarkable performance across Davies–Bouldin, Calinski–Harabasz, S_Dbw, ARI values. significantly enhanced layer resolutions accurately delineated domains breast cancer tissues, adult brain (FFPE), embryos. Conclusions serve as an essential tool bridging analysis disease pathology, offering valuable insights researchers field.

Язык: Английский

Процитировано

0
Predicting cell–cell communication by combining heterogeneous ensemble deep learning and weighted geometric mean DOI
Lihong Peng, L. C. Liu,

Liangliang Huang

и другие.

Applied Soft Computing, Год журнала: 2025, Номер unknown, С. 112839 - 112839

Опубликована: Фев. 1, 2025

Язык: Английский

Процитировано

0
MRDPDA: A multi‐Laplacian regularized deepFM model for predicting piRNA‐disease associations DOI Creative Commons
Yajun Liu, Fan Zhang, Yulian Ding

и другие.

Journal of Cellular and Molecular Medicine, Год журнала: 2024, Номер 28(17)

Опубликована: Сен. 1, 2024

Abstract PIWI‐interacting RNAs (piRNAs) are a typical class of small non‐coding RNAs, which essential for gene regulation, genome stability and so on. Accumulating studies have revealed that piRNAs significant potential as biomarkers therapeutic targets variety diseases. However current computational methods face the challenge in effectively capturing piRNA‐disease associations (PDAs) from limited data. In this study, we propose novel method, MRDPDA, predicting PDAs based on data multiple sources. Specifically, MRDPDA integrates deep factorization machine (deepFM) model with regularizations derived yet datasets, utilizing separate Laplacians instead simple average similarity network. Moreover, unified objective function to combine embedding loss about similarities is proposed ensure suitable prediction task. addition, balanced benchmark dataset piRPheno constructed autoencoder applied creating reliable negative set unlabeled dataset. Compared three latest methods, achieves best performance pirpheno terms five‐fold cross validation test independent set, case further demonstrate effectiveness MRDPDA.

Язык: Английский

Процитировано

2
Identification of drug use degree by integrating multi-modal features with dual-input deep learning method DOI
Yuxing Zhou, Xuelin Gu,

Zhen Wang

и другие.

Computer Methods in Biomechanics & Biomedical Engineering, Год журнала: 2024, Номер unknown, С. 1 - 13

Опубликована: Окт. 28, 2024

Most of studies on drug use degree are based subjective judgments without objective quantitative assessment, in this paper, a dual-input bimodal fusion algorithm is proposed to study by using electroencephalogram (EEG) and near-infrared spectroscopy (NIRS). Firstly, paper uses the optimized multi-modal TiCBnet for extracting deep encoding features signal, then fuses screens different methods, finally fused classified. The classification accuracy found be higher than that single modal, up 89.9%.

Язык: Английский

Процитировано

0
THGB: predicting ligand-receptor interactions by combining tree boosting and histogram-based gradient boosting DOI Creative Commons
Liqian Zhou, Jiao Song, Zejun Li

и другие.

Scientific Reports, Год журнала: 2024, Номер 14(1)

Опубликована: Ноя. 28, 2024

Ligand-receptor interaction (LRI) prediction has great significance in biological and medical research facilitates to infer analyze cell-to-cell communication. However, wet experiments for new LRI discovery are costly time-consuming. Here, we propose a computational model called THGB uncover LRIs. first extracts feature information of Ligand-Receptor (LR) pairs using iFeature. Next, it adopts tree boosting obtain representative LR features. Finally, devises the histogram-based gradient capture high-quality To assess performance, compared with three models (i.e., CellEnBoost, CellGiQ, CellComNet) one classical protein-protein inference PIPR. The results demonstrated that achieved best overall predictions terms six evaluation indictors precision, recall, accuracy, F1-score, AUC, AUPR). measure effect selection on prediction, was four methods PCA, NMF, LLE, TSVD). showed more appropriate select features improve prediction. We also conducted ablation study found outperformed without selection. hope is useful tool find LRIs further

Язык: Английский

Процитировано

0