deep-AMPpred: A Deep Learning Method for Identifying Antimicrobial Peptides and Their Functional Activities

Journal of Chemical Information and Modeling, Journal Year: 2025, Volume and Issue: unknown

Published: Jan. 10, 2025

Antimicrobial peptides (AMPs) are small that play an important role in disease defense. As the problem of pathogen resistance caused by misuse antibiotics intensifies, identification AMPs as alternatives to has become a hot topic. Accurately identifying using computational methods been key issue field bioinformatics recent years. Although there many machine learning-based AMP tools, most them do not focus on or only few functional activities. Predicting multiple activities antimicrobial can help discover candidate with broad-spectrum ability. We propose two-stage predictor deep-AMPpred, which first stage distinguishes from other peptides, and second solves multilabel 13 common AMP. deep-AMPpred combines ESM-2 model encode features integrates CNN, BiLSTM, CBAM models its The captures global contextual peptide sequence, while combine local feature extraction, long-term short-term dependency modeling, attention mechanisms improve performance function prediction. Experimental results demonstrate performs well accurately predicting their This confirms effectiveness capture meaningful sequence integrating deep learning for activity

Language: Английский

---

Short Text Classification Based on Enhanced Word Embedding and Hybrid Neural Networks

Applied Sciences, Journal Year: 2025, Volume and Issue: 15(9), P. 5102 - 5102

Published: May 4, 2025

In recent years, text classification has found wide application in diverse real-world scenarios. Chinese news tasks, limitations such as sparse contextual information and semantic ambiguity exist the title text. To improve performance of short classification, this paper proposes a Word2Vec-based enhanced word embedding method exhibits design dual-channel hybrid neural network architecture to effectively extract features. Specifically, we introduce novel weighting scheme, Term Frequency-Document Frequency Category-Distribution Weight (TF-IDF-CDW), where Category Distribution (CDW) reflects distribution pattern words across different categories. By pretrained Word2Vec vectors with TF-IDF-CDW concatenating them part-of-speech (POS) feature vectors, semantically enriched more discriminative are generated. Furthermore, propose model based on Gated Convolutional Neural Network (GCNN) Bidirectional Long Short-Term Memory (BiLSTM), which jointly captures local features long-range global dependencies. evaluate overall model, experiments were conducted datasets THUCNews TNews. The proposed achieved accuracies 91.85% 87.70%, respectively, outperforming several comparative models demonstrating effectiveness method.

Language: Английский

---

Efficient Agricultural Question Classification With a BERT-Enhanced DPCNN Model

IEEE Access, Journal Year: 2024, Volume and Issue: 12, P. 109255 - 109268

Published: Jan. 1, 2024

The application of big data technology in agricultural production has led to explosive growth data. accurate classification questions from vast amounts question-and-answer is currently a prominent topic text research. However, due the characteristics questions, such as short text, high specialization, and uneven sample distribution, relying on single model for feature extraction limitations. To address this issue improve performance question classification, we propose fusion BERT-DPCNN, which combines Bidirectional Encoder Representations Transformer (BERT) with Deep Pyramid Convolution Neural Network (DPCNN). Firstly, BERT pre-training captures word-level semantic information each generates hidden vectors containing sentence-level features using 12 layers transformers. Secondly, output word are input into DPCNN further extract local capture long-distance textual dependencies. Finally, verified effectiveness our self-constructed dataset. Comparative experiments demonstrate that BERT-DPCNN achieves superior results an accuracy rate 99.07%. assess its generalization performance, conducted comparison Tsinghua News Experimental show significant improvement BERT-DPCNN's datasets compared other models, meeting requirements question-answering systems.

Language: Английский

---

Contextual Semantic Embeddings Based on Transformer Models for Arabic Biomedical Questions Classification

Published: Jan. 1, 2024

Language: Английский

---

A Review on Deep Learning Based Question Answering with Natural Language Processing in Healthcare

Published: April 18, 2024

Language: Английский

---

Shared Task on NCAA 2024: Chinese Diabetes Question Classification

Communications in computer and information science, Journal Year: 2024, Volume and Issue: unknown, P. 337 - 347

Published: Sept. 21, 2024

Language: Английский

---

RBTM: A Hybrid gradient Regression-Based transformer model for biomedical question answering

Biomedical Signal Processing and Control, Journal Year: 2024, Volume and Issue: 102, P. 107325 - 107325

Published: Dec. 21, 2024

Language: Английский

---

Semantic Question Answering: Deep Learning and NoSQL Solution for the Medical Domain

2021 IEEE International Conference on Big Data (Big Data), Journal Year: 2024, Volume and Issue: unknown, P. 6486 - 6493

Published: Dec. 15, 2024

Language: Английский

---

Advanced Educational Assessments: Automated Question Classification Based on Bloom’s Cognitive Level

Published: Oct. 20, 2023

In the contemporary education system, quality of question papers plays a pivotal role in evaluating students' knowledge and comprehension. To ensure validity assessment outcomes, it is imperative to assess these papers, taking into account factors such as clarity, alignment with learning objectives, structural coherence, conformity intended educational outcomes. This study centered around development predictive model that employs Bloom's taxonomy—a framework for categorizing objectives—to gauge difficulty level questions. optimize performance, we have harnessed power Bidirectional Long Short-Term Memory Network (BiLSTM), renowned effectively preserving intricate dependencies within data. Through extensive experimentation on widely recognized datasets, our results showcased superior accuracy BiLSTM, an overall rate 80%, outperforming existing methods by substantial margin 5.44%. These findings represent significant advancement realm assessment, empowering educators advanced machine techniques more precise evaluation cognitive capabilities.

Language: Английский

---

Question Classification Method based on Self-Attention Mechanism and BiLSTM-AlexNet Model

Published: Dec. 17, 2023

Language: Английский

---