Learning Disentangled Representations of Texts with Application to Biomedical Abstracts.
Ontology highlight
ABSTRACT:
SUBMITTER: Jain S
PROVIDER: S-EPMC8136418 | biostudies-literature |
REPOSITORIES: biostudies-literature
Similar Datasets
Project description:In magnetic resonance (MR) imaging, a lack of standardization in acquisition often causes pulse sequence-based contrast variations in MR images from site to site, which impedes consistent measurements in automatic analyses. In this paper, we propose an unsupervised MR image harmonization approach, CALAMITI (Contrast Anatomy Learning and Analysis for MR Intensity Translation and Integration), which aims to alleviate contrast variations in multi-site MR imaging. Designed using information bottleneck theory, CALAMITI learns a globally disentangled latent space containing both anatomical and contrast information, which permits harmonization. In contrast to supervised harmonization methods, our approach does not need a sample population to be imaged across sites. Unlike traditional unsupervised harmonization approaches which often suffer from geometry shifts, CALAMITI better preserves anatomy by design. The proposed method is also able to adapt to a new testing site with a straightforward fine-tuning process. Experiments on MR images acquired from ten sites show that CALAMITI achieves superior performance compared with other harmonization approaches.
| S-EPMC10473284 | biostudies-literature
Project description:BackgroundThe Bacteria Biotope (BB) task is a biomedical relation extraction (RE) that aims to study the interaction between bacteria and their locations. This task is considered to pertain to fundamental knowledge in applied microbiology. Some previous investigations conducted the study by applying feature-based models; others have presented deep-learning-based models such as convolutional and recurrent neural networks used with the shortest dependency paths (SDPs). Although SDPs contain valuable and concise information, some parts of crucial information that is required to define bacterial location relationships are often neglected. Moreover, the traditional word-embedding used in previous studies may suffer from word ambiguation across linguistic contexts.ResultsHere, we present a deep learning model for biomedical RE. The model incorporates feature combinations of SDPs and full sentences with various attention mechanisms. We also used pre-trained contextual representations based on domain-specific vocabularies. To assess the model's robustness, we introduced a mean F1 score on many models using different random seeds. The experiments were conducted on the standard BB corpus in BioNLP-ST'16. Our experimental results revealed that the model performed better (in terms of both maximum and average F1 scores; 60.77% and 57.63%, respectively) compared with other existing models.ConclusionsWe demonstrated that our proposed contributions to this task can be used to extract rich lexical, syntactic, and semantic features that effectively boost the model's performance. Moreover, we analyzed the trade-off between precision and recall to choose the proper cut-off to use in real-world applications.
| S-EPMC6889521 | biostudies-literature
Project description:ObjectiveWe explored how knowledge embeddings (KEs) learned from the Unified Medical Language System (UMLS) Metathesaurus impact the quality of relation extraction on 2 diverse sets of biomedical texts.Materials and methodsTwo forms of KEs were learned for concepts and relation types from the UMLS Metathesaurus, namely lexicalized knowledge embeddings (LKEs) and unlexicalized KEs. A knowledge embedding encoder (KEE) enabled learning either LKEs or unlexicalized KEs as well as neural models capable of producing LKEs for mentions of biomedical concepts in texts and relation types that are not encoded in the UMLS Metathesaurus. This allowed us to design the relation extraction with knowledge embeddings (REKE) system, which incorporates either LKEs or unlexicalized KEs produced for relation types of interest and their arguments.ResultsThe incorporation of either LKEs or unlexicalized KE in REKE advances the state of the art in relation extraction on 2 relation extraction datasets: the 2010 i2b2/VA dataset and the 2013 Drug-Drug Interaction Extraction Challenge corpus. Moreover, the impact of LKEs is superior, achieving F1 scores of 78.2 and 82.0, respectively.DiscussionREKE not only highlights the importance of incorporating knowledge encoded in the UMLS Metathesaurus in a novel way, through 2 possible forms of KEs, but it also showcases the subtleties of incorporating KEs in relation extraction systems.ConclusionsIncorporating LKEs informed by the UMLS Metathesaurus in a relation extraction system operating on biomedical texts shows significant promise. We present the REKE system, which establishes new state-of-the-art results for relation extraction on 2 datasets when using LKEs.
| S-EPMC7647370 | biostudies-literature
Project description:Deep generative models such as variational autoencoders (VAEs) and generative adversarial networks (GANs) generate and manipulate high-dimensional images. We systematically assess the complementary strengths and weaknesses of these models on single-cell gene expression data. We also develop MichiGAN, a novel neural network that combines the strengths of VAEs and GANs to sample from disentangled representations without sacrificing data generation quality. We learn disentangled representations of three large single-cell RNA-seq datasets and use MichiGAN to sample from these representations. MichiGAN allows us to manipulate semantically distinct aspects of cellular identity and predict single-cell gene expression response to drug treatment.
| S-EPMC8139054 | biostudies-literature
Project description:SummaryAlthough the publication rate of the biomedical literature has been growing steadily during the last decades, the accessibility of pertinent research publications for biologist and medical practitioners remains a challenge. This article describes Thalia, which is a semantic search engine that can recognize eight different types of concepts occurring in biomedical abstracts. Thalia is available via a web-based interface or a RESTful API. A key aspect of our search engine is that it is updated from PubMed on a daily basis. We describe here the main building blocks of our tool as well as an evaluation of the retrieval capabilities of Thalia in the context of a precision medicine dataset.Availability and implementationThalia is available at http://nactem.ac.uk/Thalia_BI/.Supplementary informationSupplementary data are available at Bioinformatics online.
| S-EPMC6513154 | biostudies-literature
Project description:Recent developments of single cell RNA-sequencing technologies lead to the exponential growth of single cell sequencing datasets across different conditions. Combining these datasets helps to better understand cellular identity and function. However, it is challenging to integrate different datasets from different laboratories or technologies due to batch effect, which are interspersed with biological variances. To overcome this problem, we have proposed Single Cell Integration by Disentangled Representation Learning (SCIDRL), a domain adaption-based method, to learn low-dimensional representations invariant to batch effect. This method can efficiently remove batch effect while retaining cell type purity. We applied it to thirteen diverse simulated and real datasets. Benchmark results show that SCIDRL outperforms other methods in most cases and exhibits excellent performances in two common situations: (i) effective integration of batch-shared rare cell types and preservation of batch-specific rare cell types; (ii) reliable integration of datasets with different cell compositions. This demonstrates SCIDRL will offer a valuable tool for researchers to decode the enigma of cell heterogeneity.
| S-EPMC8788944 | biostudies-literature
Project description: Not available
| S-EPMC6985892 | biostudies-literature
Project description:MotivationRecent neural approaches on event extraction from text mainly focus on flat events in general domain, while there are less attempts to detect nested and overlapping events. These existing systems are built on given entities and they depend on external syntactic tools.ResultsWe propose an end-to-end neural nested event extraction model named DeepEventMine that extracts multiple overlapping directed acyclic graph structures from a raw sentence. On the top of the bidirectional encoder representations from transformers model, our model detects nested entities and triggers, roles, nested events and their modifications in an end-to-end manner without any syntactic tools. Our DeepEventMine model achieves the new state-of-the-art performance on seven biomedical nested event extraction tasks. Even when gold entities are unavailable, our model can detect events from raw text with promising performance.Availability and implementationOur codes and models to reproduce the results are available at: https://github.com/aistairc/DeepEventMine.Supplementary informationSupplementary data are available at Bioinformatics online.
| S-EPMC7750964 | biostudies-literature
Project description:BackgroundThe realm of pathological entities can be subdivided into pathological dispositions, pathological processes, and pathological structures. The latter are the bearer of dispositions, which can then be realized by their manifestations - pathologic processes. Despite its ontological soundness, implementing this model via purpose-oriented domain ontologies will likely require considerable effort, both in ontology construction and maintenance, which constitutes a considerable problem for SNOMED CT, presently the largest biomedical ontology.ResultsWe describe an ontology design pattern which allows ontologists to make assertions that blur the distinctions between dispositions, processes, and structures until necessary. Based on the domain upper-level ontology BioTop, it permits ascriptions of location and participation in the definition of pathological phenomena even without an ontological commitment to a distinction between these three categories. An analysis of SNOMED CT revealed that numerous classes in the findings/disease hierarchy are ambiguous with respect to process vs. disposition. Here our proposed approach can easily be applied to create unambiguous classes. No ambiguities could be defined regarding the distinction of structure and non-structure classes, but here we have found problematic duplications.ConclusionsWe defend a judicious use of disjunctive, and therefore ambiguous, classes in biomedical ontologies during the process of ontology construction and in the practice of ontology application. The use of these classes is permitted to span across several top-level categories, provided it contributes to ontology simplification and supports the intended reasoning scenarios.
| S-EPMC3102895 | biostudies-literature
Project description:The ever accelerating pace of biomedical research results in corresponding acceleration in the volume of biomedical literature created. Since new research builds upon existing knowledge, the rate of increase in the available knowledge encoded in biomedical literature makes the easy access to that implicit knowledge more vital over time. Toward the goal of making implicit knowledge in the biomedical literature easily accessible to biomedical researchers, we introduce a question answering system called Bio-AnswerFinder. Bio-AnswerFinder uses a weighted-relaxed word mover's distance based similarity on word/phrase embeddings learned from PubMed abstracts to rank answers after question focus entity type filtering. Our approach retrieves relevant documents iteratively via enhanced keyword queries from a traditional search engine. To improve document retrieval performance, we introduced a supervised long short term memory neural network to select keywords from the question to facilitate iterative keyword search. Our unsupervised baseline system achieves a mean reciprocal rank score of 0.46 and Precision@1 of 0.32 on 936 questions from BioASQ. The answer sentences are further ranked by a fine-tuned bidirectional encoder representation from transformers (BERT) classifier trained using 100 answer candidate sentences per question for 492 BioASQ questions. To test ranking performance, we report a blind test on 100 questions that three independent annotators scored. These experts preferred BERT based reranking with 7% improvement on MRR and 13% improvement on Precision@1 scores on average.
| S-EPMC7053013 | biostudies-literature