Integrated convolution and self-attention for improving peptide toxicity prediction

Bioinformatics, Год журнала: 2024, Номер 40(5)

Опубликована: Май 1, 2024

Abstract Motivation Peptides are promising agents for the treatment of a variety diseases due to their specificity and efficacy. However, development peptide-based drugs is often hindered by potential toxicity peptides, which poses significant barrier clinical application. Traditional experimental methods evaluating peptide time-consuming costly, making process inefficient. Therefore, there an urgent need computational tools specifically designed predict accurately rapidly, facilitating identification safe candidates drug development. Results We provide here novel approach, CAPTP, leverages power convolutional self-attention enhance prediction from amino acid sequences. CAPTP demonstrates outstanding performance, achieving Matthews correlation coefficient approximately 0.82 in both cross-validation settings on independent test datasets. This performance surpasses that existing state-of-the-art predictors. Importantly, maintains its robustness generalizability even when dealing with data imbalances. Further analysis reveals certain sequential patterns, particularly head central regions crucial determining toxicity. insight can significantly inform guide design safer drugs. Availability implementation The source code freely available at https://github.com/jiaoshihu/CAPTP.

Язык: Английский

---

Machine Learning‐Enabled Drug‐Induced Toxicity Prediction

Advanced Science, Год журнала: 2025, Номер unknown

Опубликована: Фев. 3, 2025

Abstract Unexpected toxicity has become a significant obstacle to drug candidate development, accounting for 30% of discovery failures. Traditional assessment through animal testing is costly and time‐consuming. Big data artificial intelligence (AI), especially machine learning (ML), are robustly contributing innovation progress in toxicology research. However, the optimal AI model different types usually varies, making it essential conduct comparative analyses methods across domains. The diverse sources also pose challenges researchers focusing on specific studies. In this review, 10 categories drug‐induced examined, summarizing characteristics applicable ML models, including both predictive interpretable algorithms, striking balance between breadth depth. Key databases tools used prediction highlighted, toxicology, chemical, multi‐omics, benchmark databases, organized by their focus function clarify roles prediction. Finally, strategies turn into opportunities analyzed discussed. This review may provide with valuable reference understanding utilizing available resources bridge mechanistic insights, further advance application drugs‐induced

Язык: Английский

---

GSAML-DTA: An interpretable drug-target binding affinity prediction model based on graph neural networks with self-attention mechanism and mutual information

Computers in Biology and Medicine, Год журнала: 2022, Номер 150, С. 106145 - 106145

Опубликована: Окт. 4, 2022

Язык: Английский

---

MolFPG: Multi-level fingerprint-based Graph Transformer for accurate and robust drug toxicity prediction

Computers in Biology and Medicine, Год журнала: 2023, Номер 164, С. 106904 - 106904

Опубликована: Май 15, 2023

Язык: Английский

---

DeepAFP: An effective computational framework for identifying antifungal peptides based on deep learning

Protein Science, Год журнала: 2023, Номер 32(10)

Опубликована: Авг. 19, 2023

Abstract Fungal infections have become a significant global health issue, affecting millions worldwide. Antifungal peptides (AFPs) emerged as promising alternative to conventional antifungal drugs due their low toxicity and propensity for inducing resistance. In this study, we developed deep learning‐based framework called DeepAFP efficiently identify AFPs. fully leverages mines composition information, evolutionary physicochemical properties of by employing combined kernels from multiple branches convolutional neural network with bi‐directional long short‐term memory layers. addition, integrates transfer learning strategy obtain efficient representations improving model performance. demonstrates strong predictive ability on carefully curated datasets, yielding an accuracy 93.29% F1‐score 93.45% the DeepAFP‐Main dataset. The experimental results show that outperforms existing AFP prediction tools, achieving state‐of‐the‐art Finally, provide downloadable tool meet demands large‐scale facilitate usage our public or other researchers. Our can accurately AFPs in short time without requiring human material resources, hence accelerate development well contribute treatment fungal infections. Furthermore, method new perspectives biological sequence analysis tasks.

Язык: Английский

---

Accelerating the Discovery of Anticancer Peptides through Deep Forest Architecture with Deep Graphical Representation

International Journal of Molecular Sciences, Год журнала: 2023, Номер 24(5), С. 4328 - 4328

Опубликована: Фев. 21, 2023

Cancer is one of the leading diseases threatening human life and health worldwide. Peptide-based therapies have attracted much attention in recent years. Therefore, precise prediction anticancer peptides (ACPs) crucial for discovering designing novel cancer treatments. In this study, we proposed a machine learning framework (GRDF) that incorporates deep graphical representation forest architecture identifying ACPs. Specifically, GRDF extracts features based on physicochemical properties integrates their evolutionary information along with binary profiles constructing models. Moreover, employ algorithm, which adopts layer-by-layer cascade similar to neural networks, enabling excellent performance small datasets but without complicated tuning hyperparameters. The experiment shows exhibits state-of-the-art two elaborate (Set 1 Set 2), achieving 77.12% accuracy 77.54% F1-score 1, as well 94.10% 94.15% 2, exceeding existing ACP methods. Our models exhibit greater robustness than baseline algorithms commonly used other sequence analysis tasks. addition, well-interpretable, researchers better understand peptide sequences. promising results demonstrate remarkably effective presented study could assist facilitating discovery contribute developing

Язык: Английский

---

Reshaping the discovery of self-assembling peptides with generative AI guided by hybrid deep learning

Nature Machine Intelligence, Год журнала: 2024, Номер unknown

Опубликована: Ноя. 19, 2024

Язык: Английский

---

Improved Prediction Model of Protein and Peptide Toxicity by Integrating Channel Attention into a Convolutional Neural Network and Gated Recurrent Units

ACS Omega, Год журнала: 2022, Номер 7(44), С. 40569 - 40577

Опубликована: Окт. 27, 2022

In recent times, the importance of peptides in biomedical domain has received increasing concern terms their effect on multiple disease treatments. However, before successful large-scale implementation industry, accurate identification peptide toxicity is a vital prerequisite. The existing computational methods have reached good results from prediction, and we present an improved model based different deep learning architectures. modification mainly focuses two aspects: sequence encoding variational information bottlenecks. Consequently, one our modified plans shows obvious increase sensitivity, while rest show performance meanwhile adding novelty prediction domain. detail, best could achieve accuracy 97.38 95.03% protein predictions, respectively. was superior to previous predictors same datasets.

Язык: Английский

---

Invited review: Camel milk–derived bioactive peptides and diabetes—Molecular view and perspectives

Journal of Dairy Science, Год журнала: 2023, Номер 107(2), С. 649 - 668

Опубликована: Сен. 13, 2023

In dairy science, camel milk (CM) constitutes a center of interest for scientists due to its known beneficial effect on diabetes as demonstrated in many vitro, vivo, and clinical studies trials. Overall, CM had positive effects various parameters related glucose transport metabolism well the structural functional properties pancreatic β-cells insulin secretion. Thus, consumption may help manage diabetes; however, such recommendation will become rationale clinically conceivable only if exact molecular mechanisms pathways involved at cellular levels are understood. Moreover, application an alternative antidiabetic tool first require identification bioactive molecules behind properties. this review, we describe advances our knowledge reported be managing using different vitro vivo models. This mainly includes controlling (1) receptor signaling uptake, (2) β-cell structure function, (3) activity key metabolic enzymes metabolism. described current status CM-derived peptides their structure-activity relationship study characterization context markers diabetes. Such overview not enrich scientific plausible mode action but should ultimately rationalize claim potential against pave way toward new directions ideas developing generation products taking benefits from chemical composition CM.

Язык: Английский

---

Structure‐aware deep learning model for peptide toxicity prediction

Protein Science, Год журнала: 2024, Номер 33(7)

Опубликована: Июнь 22, 2024

Antimicrobial resistance is a critical public health concern, necessitating the exploration of alternative treatments. While antimicrobial peptides (AMPs) show promise, assessing their toxicity using traditional wet lab methods both time-consuming and costly. We introduce tAMPer, novel multi-modal deep learning model designed to predict peptide by integrating underlying amino acid sequence composition three-dimensional structure peptides. tAMPer adopts graph-based representation for peptides, encoding ColabFold-predicted structures, where nodes represent acids edges spatial interactions. Structural features are extracted graph neural networks, recurrent networks capture sequential dependencies. tAMPer's performance was assessed on publicly available protein benchmark an AMP hemolysis data we generated. On latter, achieves F1-score 68.7%, outperforming second-best method 23.4%. benchmark, exhibited improvement over 3.0% in compared current state-of-the-art methods. anticipate accelerate discovery development reducing reliance laborious screening experiments.

Язык: Английский

---