Advancements and implications of artificial intelligence for early detection, diagnosis and tailored treatment of cancer

Seminars in Oncology, Год журнала: 2025, Номер 52(3), С. 152349 - 152349

Опубликована: Май 8, 2025

Язык: Английский

---

Genome language modeling (GLM): a beginner’s cheat sheet

Biology Methods and Protocols, Год журнала: 2025, Номер 10(1)

Опубликована: Янв. 1, 2025

Abstract Integrating genomics with diverse data modalities has the potential to revolutionize personalized medicine. However, this integration poses significant challenges due fundamental differences in types and structures. The vast size of genome necessitates transformation into a condensed representation containing key biomarkers relevant features ensure interoperability other modalities. This commentary explores both conventional state-of-the-art approaches language modeling (GLM), focus on representing extracting meaningful from genomic sequences. We latest trends applying techniques sequence data, treating it as text modality. Effective feature extraction is essential enabling machine learning models effectively analyze large datasets, particularly within multimodal frameworks. first provide step-by-step guide various preprocessing tokenization techniques. Then we explore methods for tokens using frequency, embedding, neural network-based approaches. In end, discuss (ML) applications genomics, focusing classification, regression, processing algorithms, integration. Additionally, role GLM functional annotation, emphasizing how advanced ML models, such Bidirectional encoder representations transformers, enhance interpretation data. To best our knowledge, compile end-to-end analytic convert complex biologically interpretable information GLM, thereby facilitating development novel data-driven hypotheses.

Язык: Английский

---

Transformer-based models for uncovering genetic mutations in cancerous and non-cancerous genomes

Gene, Год журнала: 2025, Номер 963, С. 149460 - 149460

Опубликована: Май 27, 2025

Язык: Английский

---

Deep-GenMut: Automated genetic mutation classification in oncology: A deep learning comparative study

Heliyon, Год журнала: 2024, Номер 10(11), С. e32279 - e32279

Опубликована: Май 31, 2024

Early cancer detection and treatment depend on the discovery of specific genes that cause cancer. The classification genetic mutations was initially done manually. However, this process relies pathologists can be a time-consuming task. Therefore, to improve precision clinical interpretation, researchers have developed computational algorithms leverage next-generation sequencing technologies for automated mutation analysis. This paper utilized four deep learning models with training collections biomedical texts. These comprise bidirectional encoder representations from transformers Biomedical text mining (BioBERT), specialized language model implemented biological contexts. Impressive results in multiple tasks, including classification, inference, question answering, obtained by simply adding an extra layer BioBERT model. Moreover, (BERT), long short-term memory (LSTM), LSTM (BiLSTM) been leveraged produce very good categorizing based textual evidence. dataset used work created Memorial Sloan Kettering Cancer Center (MSKCC), which contains several mutations. Furthermore, poses major challenge Kaggle research prediction competitions. In carrying out work, three challenges were identified: enormous length, biased representation data, repeated data instances. Based commonly evaluation metrics, experimental show outperforms other F1 score 0.87 0.850 MCC, considered as improved performance compared similar literature 0.70 achieved BERT

Язык: Английский

---

Authenticity in authorship: the Writer’s Integrity framework for verifying human-generated text

Ethics and Information Technology, Год журнала: 2024, Номер 26(3)

Опубликована: Сен. 1, 2024

Язык: Английский

---

Comparative Analysis of LSTM and Ensemble LSTM Approaches for Gene Mutation Classification in Cancer

Опубликована: Дек. 14, 2023

In this study, we present an in-depth comparison of five different deep learning approaches for the classification gene mutations based on a dataset provided by Kaggle competition "Personalized Medicine: Redefining Cancer Treatment." The models compared include Long Short-Term Memory (LSTM) model, ensemble LSTM and Bidirectional (BiLSTM), 1-Dimensional Convolutional Neural Network (1D-CNN), Gated Recurrent Unit (GRU), multi-ensemble model combining LSTM, BiLSTM, 1D-CNN, GRU. These were evaluated several metrics including accuracy, precision, recall, F1 score, mean squared error (MSE) both training validation sets. Among all models, + 1D-CNN demonstrated superior performance set while also being most time-efficient to train. results contribute growing body research in field personalized medicine highlight efficacy mutations, which could play vital role future cancer treatment strategies.

Язык: Английский

---

Automatic Extractive Summarization using GAN Boosted by DistilBERT Word Embedding and Transductive Learning

International Journal of Advanced Computer Science and Applications, Год журнала: 2023, Номер 14(11)

Опубликована: Янв. 1, 2023

Text summarization is crucial in diverse fields such as engineering and healthcare, greatly enhancing time cost efficiency. This study introduces an innovative extractive text approach utilizing a Generative Adversarial Network (GAN), Transductive Long Short-Term Memory (TLSTM), DistilBERT word embedding. DistilBERT, streamlined BERT variant, offers significant size reduction (approximately 40%), while maintaining 97% of language comprehension capabilities achieving 60% speed increase. These benefits are realized through knowledge distillation during pre-training. Our methodology uses GANs, consisting the generator discriminator networks, built primarily using TLSTM - expert at decoding temporal nuances timeseries prediction. For more effective model fitting, transductive learning employed, assigning higher weights to samples nearer test point. The evaluates probability each sentence for inclusion summary, critically examines generated summary. reciprocal relationship fosters dynamic iterative process, generating top-tier summaries. To train efficiently, unique loss function proposed, incorporating multiple factors generator’s output, actual document summaries, artificially created strategy motivates experiment with combinations, summaries that meet high-quality coherence standards. model’s effectiveness was tested on widely accepted CNN/Daily Mail dataset, benchmark tasks. According ROUGE metric, our experiments demonstrate outperforms existing models terms quality

Язык: Английский

---

Construction of Big Data Information Security Protection System in Industrial Internet Environment

Applied Mathematics and Nonlinear Sciences, Год журнала: 2024, Номер 9(1)

Опубликована: Янв. 1, 2024

Abstract With the continuous development and integration of information technology industrialization-related technologies, industrial Internet control system security attacks occur frequently, it is more important to build an protection system. This study focuses on research improvement from two aspects access intrusion prevention designs strategy based homomorphic encryption algorithm Hyper Elliptic Curve Cryptosystem (HCC) key splitting threshold. Meanwhile, convolutional neural network, two-way gating loop unit, multi-head attention mechanism are integrated construct CMAG detection model. The model applied analyzed. decryption times this paper’s both relatively smooth, with average time consumption about 1.93ms 0.46ms, respectively, significantly better than other algorithms increase in number bits. throughput 13.68 KB/s, which approximately 2 times, 19 29 higher GM, ElGamal, Paillier algorithms, respectively. cannot match its rate during decryption. has accuracy 99.14%, that models, checking accuracy, recall, F1-Score 0.9889, 0.9783, 0.9834, 1.25%-5.16%, 4.31%-7.19%, 3.32%, compared three algorithms. 7.19% 3.32%-6.76%, paper great practical significance for construction optimization a big data environment.

Язык: Английский

---

Anomaly Detection in Embryo Development and Morphology Using Medical Computer Vision-Aided Swin Transformer with Boosted Dipper-Throated Optimization Algorithm

Bioengineering, Год журнала: 2024, Номер 11(10), С. 1044 - 1044

Опубликована: Окт. 18, 2024

Infertility affects a significant number of humans. A supported reproduction technology was verified to ease infertility problems. In vitro fertilization (IVF) is one the best choices, and its success relies on preference for higher-quality embryo transmission. These have been normally completed physically by testing embryos in microscope. The traditional morphological calculation shows predictable disadvantages, including effort- time-consuming expected risks bias related individual estimations specific embryologists. Different computer vision (CV) artificial intelligence (AI) techniques devices recently applied fertility hospitals improve efficacy. AI addresses imitation intellectual performance capability technologies simulate cognitive learning, thinking, problem-solving typically Deep learning (DL) machine (ML) are advanced algorithms various fields considered main future human assistant technology. This study presents an Embryo Development Morphology Using Computer Vision-Aided Swin Transformer with Boosted Dipper-Throated Optimization (EDMCV-STBDTO) technique. EDMCV-STBDTO technique aims accurately efficiently detect development, which critical improving treatments advancing developmental biology using medical CV techniques. Primarily, method performs image preprocessing bilateral filter (BF) model remove noise. Next, swin transformer implemented feature extraction employs variational autoencoder (VAE) classify development. Finally, hyperparameter selection VAE boosted dipper-throated optimization (BDTO) efficiency validated comprehensive studies benchmark dataset. experimental result that better than recent

Язык: Английский

---

TURKISH SIGN LANGUAGE EXPRESSIONS RECOGNITION USING DEEP LEARNING AND LANDMARK DATA

Mugla Journal of Science and Technology, Год журнала: 2024, Номер unknown

Опубликована: Ноя. 17, 2024

Sign language is a vital communication tool for hearing-impaired individuals to express their thoughts and emotions. Turkish Language (TSL) based on hand gestures, facial expressions, body movements. In this study, deep learning models were developed recognize 41 commonly used TSL expressions. An original dataset was created using the Media Pipe Holistic framework capture 3D landmarks of hand, face, The study trained evaluated GRU, LSTM, Bi-LSTM models, as well hybrid architectures such CNN+GRU, GRU+LSTM, GRU+Bi-LSTM. training hold-out validation method used. 80% allocated 20% testing. Additionally, data validation. Among Deep Learning CNN+GRU model achieved highest accuracy rate 96.72%, outperforming similar studies in literature. Our results demonstrate that techniques can effectively classify with combination showing particularly high performance. Future work will focus expanding developing real-time recognition systems incorporate both skeleton images landmarks.

Язык: Английский

---