PINNED: Identifying Characteristics of Druggable Human Proteins Using an Interpretable Neural Network DOI Creative Commons
Michael Cunningham,

Danielle Pins,

Zoltán Dezső

et al.

Published: March 30, 2023

The identification of human proteins that are amenable to pharmacologic modulation without significant off-target effects remains an important unsolved challenge. Computational methods have been devised identify features which distinguish between “druggable” and “undruggable” proteins, finding protein sequence, tissue cellular localization, biological role, position in the protein-protein interaction network all discriminant factors. However, many prior efforts automate assessment druggability suffer from low performance or poor interpretability. We developed a neural network-based machine learning model capable generating sub-scores based on each four distinct categories, combining them form overall score. achieves excellent separating drugged undrugged proteome, with area under receiver operating characteristic (AUC) 0.95. Our use multiple allows potential targets interest contributors druggability, leading more interpretable holistic novel targets.

Language: Английский

DPI_CDF: druggable protein identifier using cascade deep forest DOI Creative Commons
Muhammad Arif, Fang Ge, Ali Ghulam

et al.

BMC Bioinformatics, Journal Year: 2024, Volume and Issue: 25(1)

Published: April 5, 2024

Abstract Background Drug targets in living beings perform pivotal roles the discovery of potential drugs. Conventional wet-lab characterization drug is although accurate but generally expensive, slow, and resource intensive. Therefore, computational methods are highly desirable as an alternative to expedite large-scale identification druggable proteins (DPs); however, existing silico predictor’s performance still not satisfactory. Methods In this study, we developed a novel deep learning-based model DPI_CDF for predicting DPs based on protein sequence only. utilizes evolutionary-based (i.e., histograms oriented gradients position-specific scoring matrix), physiochemical-based component representation), compositional-based normalized qualitative characteristic) properties generate features. Then hierarchical forest fuses these three encoding schemes build proposed DPI_CDF. Results The empirical outcomes 10-fold cross-validation demonstrate that achieved 99.13 % accuracy 0.982 Matthew’s-correlation-coefficient (MCC) training dataset. generalization power trained further examined independent dataset 95.01% maximum 0.900 MCC. When compared current state-of-the-art methods, improves terms by 4.27% 4.31% testing datasets, respectively. We believe, will support research community identify escalate process. Availability benchmark datasets source codes available GitHub: http://github.com/Muhammad-Arif-NUST/DPI_CDF .

Language: Английский

Citations

11
IP-GCN: A Deep Learning Model for Prediction of Insulin using Graph Convolutional Network for Diabetes Drug Design DOI
Farman Ali,

Majdi Khalid,

Abdullah Almuhaimeed

et al.

Journal of Computational Science, Journal Year: 2024, Volume and Issue: 81, P. 102388 - 102388

Published: July 14, 2024

Language: Английский

Citations

8
A stacking ensemble model for predicting soil organic carbon content based on visible and near-infrared spectroscopy DOI
Ke Tang, Xing Zhao,

Zong Xu

et al.

Infrared Physics & Technology, Journal Year: 2024, Volume and Issue: 140, P. 105404 - 105404

Published: June 15, 2024

Language: Английский

Citations

6
Machine learning-based model for accurate identification of druggable proteins using light extreme gradient boosting DOI
Omar Alghushairy, Farman Ali, Wajdi Alghamdi

et al.

Journal of Biomolecular Structure and Dynamics, Journal Year: 2023, Volume and Issue: 42(22), P. 12330 - 12341

Published: Oct. 18, 2023

AbstractThe identification of druggable proteins (DPs) is significant for the development new drugs, personalized medicine, understanding disease mechanisms, drug repurposing, and economic benefits. By identifying targets, researchers can develop therapies a range diseases, leading to better patient outcomes. Identification DPs by machine learning strategies more efficient cost-effective than conventional methods. In this study, computational predictor, namely Drug-LXGB, introduced enhance DPs. Features are discovered composition, transition, distribution (CTD), composition K-spaced amino acid pair (CKSAAP), pseudo-position-specific scoring matrix (PsePSSM), novel descriptor, called multi-block pseudo (MB-PseAAC). The dimensions CTD, CKSAAP, PsePSSM, MB-PseAAC integrated utilized sequential forward selection as feature algorithm. best characteristics provided random forest, extreme gradient boosting, light eXtreme boosting (LXGB). predictive analysis these methods measured via 10-fold cross-validation. LXGB-based model secures highest results other existing predictors. Our protocol will perform an active role in designing drugs would be fruitful explore potential target. This study help capture universal view target.Communicated Ramaswamy H. SarmaKeywords: Druggable proteinslight boostingmachine AcknowledgmentsThe authors gratefully acknowledge technical financial support Ministry Education King Abdulaziz University, DSR, Jeddah, Saudi Arabia.Disclosure statementThe have no competing interest.Additional informationFundingThis research work was supported Institutional Fund Projects under grant number IFPIP: 1396-611-1443.

Language: Английский

Citations

16
Identification of potential novel therapeutic drug target against Elizabethkingia anophelis by integrative pan and subtractive genomic analysis: An in silico approach DOI

Parth Sarker,

Arnob Mitro,

Hammadul Hoque

et al.

Computers in Biology and Medicine, Journal Year: 2023, Volume and Issue: 165, P. 107436 - 107436

Published: Aug. 30, 2023

Language: Английский

Citations

11
PMPred-AE: a computational model for the detection and interpretation of pathological myopia based on artificial intelligence DOI Creative Commons

Hongqi Zhang,

Muhammad Arif, Maha A. Thafar

et al.

Frontiers in Medicine, Journal Year: 2025, Volume and Issue: 12

Published: March 13, 2025

Introduction Pathological myopia (PM) is a serious visual impairment that may lead to irreversible damage or even blindness. Timely diagnosis and effective management of PM are great significance. Given the increasing number cases worldwide, there an urgent need develop automated, accurate, highly interpretable diagnostic technology. Methods We proposed computational model called PMPred-AE based on EfficientNetV2-L with attention mechanism optimization. In addition, Gradient-weighted class activation mapping (Grad-CAM) technology was used provide intuitive interpretation for model’s decision-making process. Results The experimental results demonstrated achieved excellent performance in automatically detecting PM, accuracies 98.50, 98.25, 97.25% training, validation, test datasets, respectively. can focus specific areas image when making detection decisions. Discussion developed capable reliably providing accurate detection. Grad-CAM also process model. This approach provides healthcare professionals tool AI

Language: Английский

Citations

0
M3S-GRPred: a novel ensemble learning approach for the interpretable prediction of glucocorticoid receptor antagonists using a multi-step stacking strategy DOI Creative Commons
Nalini Schaduangrat, Hathaichanok Chuntakaruk, Thanyada Rungrotmongkol

et al.

BMC Bioinformatics, Journal Year: 2025, Volume and Issue: 26(1)

Published: April 30, 2025

Accelerating drug discovery for glucocorticoid receptor (GR)-related disorders, including innovative machine learning (ML)-based approaches, holds promise in advancing therapeutic development, optimizing treatment efficacy, and mitigating adverse effects. While experimental methods can accurately identify GR antagonists, they are often not cost-effective large-scale discovery. Thus, computational approaches leveraging SMILES information precise silico identification of antagonists crucial, enabling efficient scalable Here, we develop a new ensemble approach using multi-step stacking strategy (M3S), termed M3S-GRPred, aimed at rapidly discovering novel antagonists. To the best our knowledge, M3S-GRPred is first SMILES-based predictor designed to without use 3D structural information. In constructed different balanced subsets an under-sampling approach. Using these subsets, explored evaluated heterogeneous base-classifiers trained with variety feature descriptors coupled popular ML algorithms. Finally, was by integrating probabilistic from selected derived two-step selection technique. Our comparative experiments demonstrate that precisely effectively address imbalanced dataset. Compared traditional classifiers, attained superior performance terms both training independent test datasets. Additionally, applied potential among FDA-approved drugs confirmed through molecular docking, followed detailed MD simulation studies repurposing Cushing's syndrome. We anticipate will serve as screening tool vast libraries unknown compounds manner.

Language: Английский

Citations

0
Collective in-silico and in-vitro evaluation indicate natural phenolics as a potential therapeutic candidate targeting antimicrobial-resistant genes of Helicobacter pylori DOI
Neha Jaiswal, Meenakshi Kandpal, Hem Chandra Jha

et al.

International Journal of Biological Macromolecules, Journal Year: 2025, Volume and Issue: unknown, P. 142197 - 142197

Published: March 1, 2025

Language: Английский

Citations

0
Repositioning of clinically approved drug Bazi Bushen capsule for treatment of Aizheimer's disease using network pharmacology approach and in vitro experimental validation DOI Creative Commons
Tongxing Wang, Meng Chen, Huixin Li

et al.

Heliyon, Journal Year: 2023, Volume and Issue: 9(7), P. e17603 - e17603

Published: June 27, 2023

AimsTo explore the new indications and key mechanism of Bazi Bushen capsule (BZBS) by network pharmacology in vitro experiment.MethodsThe ingredients library BZBS was constructed retrieving multiple TCM databases. The potential target profiles components were predicted prediction algorithms based on different principles, validated using known activity data. spectrum with high reliability screened considering source targets node degree compound-target (C-T) network. Subsequently, for disease ontology (DO) enrichment analysis initially GO KEGG pathway analysis. Furthermore, sets acting AD signaling identified intersection Based STRING database, PPI their calculated. Two Alzheimer's (AD) cell models, BV-2 SH-SY5Y, used to preliminarily verify anti-AD efficacy vitro.ResultsIn total, 1499 non-repeated obtained from 16 herbs formula, 1320 confidence predicted. Disease results strongly suggested that formula has be treatment AD. provide a preliminary verification this point. Among them, 113 functional belong pathway. A containing 1051 edges constructed. In experiments showed could significantly reduce release TNF-α IL-6 expression COX-2 PSEN1 Aβ25-35-induced cells, which may related regulation ERK1/2/NF-κB reduced apoptosis rate Aβ25-35 induced SH-SY5Y increased mitochondrial membrane potential, Caspase3 active fragment PSEN1, IDE. This GSK-3β/β-catenin pathway.ConclusionsBZBS use AD, is achieved through ERK1/2, NF-κB pathways, technology feasible drug repurposing strategy reposition clinical approved action. study lays foundation subsequent in-depth provides basis its application

Language: Английский

Citations

7
In silico protein function prediction: the rise of machine learning-based approaches DOI Creative Commons
Jiaxiao Chen, Zhonghui Gu, Luhua Lai

et al.

Medical Review, Journal Year: 2023, Volume and Issue: 3(6), P. 487 - 510

Published: Nov. 28, 2023

Abstract Proteins function as integral actors in essential life processes, rendering the realm of protein research a fundamental domain that possesses potential to propel advancements pharmaceuticals and disease investigation. Within context research, an imperious demand arises uncover functionalities untangle intricate mechanistic underpinnings. Due exorbitant costs limited throughput inherent experimental investigations, computational models offer promising alternative accelerate annotation. In recent years, pre-training have exhibited noteworthy advancement across multiple prediction tasks. This highlights notable prospect for effectively tackling downstream task associated with prediction. this review, we elucidate historical evolution paradigms methods predicting function. Subsequently, summarize progress molecule representation well feature extraction techniques. Furthermore, assess performance machine learning-based algorithms various objectives prediction, thereby offering comprehensive perspective on within field.

Language: Английский

Citations

7