Knowledge-guided Pretext Learning for Utero-placental Interface Detection.
Ontology highlight
ABSTRACT:
SUBMITTER: Qi H
PROVIDER: S-EPMC7610196 | biostudies-literature | 2020 Oct
REPOSITORIES: biostudies-literature
Publications
Knowledge-guided Pretext Learning for Utero-placental Interface Detection.
Qi Huan H Collins Sally S Noble J Alison JA
Medical image computing and computer-assisted intervention : MICCAI ... International Conference on Medical Image Computing and Computer-Assisted Intervention 20200929
Modern machine learning systems, such as convolutional neural networks rely on a rich collection of training data to learn discriminative representations. In many medical imaging applications, unfortunately, collecting a large set of well-annotated data is prohibitively expensive. To overcome data shortage and facilitate representation learning, we develop Knowledge-guided Pretext Learning (KPL) that learns anatomy-related image representations in a pretext task under the guidance of knowledge f ...[more]
Similar Datasets
Project description:What we understand about early stages of placentation in cattle is based on an elegant series of electron microscopic images that provide exquisite detail, but limited appreciation for the microanatomy across the utero-placental interface. In order to achieve a global perspective on the histology of bovine placentation during critical early stages of gestation, i.e., days 21, 31, 40, and 67, we performed immunohistochemistry to detect cell-specific expression of pregnancy-associated glycoprotein (PAG), cytokeratin, epithelial (E)-cadherin, and serine hydroxymethyltransferase 2 (SHMT2) at the intact utero-placental interface. Key findings from the immunohistochemical analyses are that there are: (1) PAG-positive cells with a single nucleus within the uterine luminal epithelial (LE) cells; (2) PAG-positive cells with two nuclei in the LE; (3) PAG-positive syncytial cells with more than three nuclei in the LE; (4) LE cells that are dissociated from one another and dissociated from the basement membrane in regions of syncytialization within the LE layer; (5) replacement of the mononuclear LE with a multi-layer thick population of PAG-positive cells invading into the uterine stroma of caruncles, but not into the stroma of intercaruncular endometrium; and (6) PAG-, E-cadherin- and SHMT2-positive mononuclear cells at the leading edge of developing cotyledonary villi that eventually represent the majority of the epithelial surface separating caruncular stroma from cotyledonary stroma. Finally, the utero-placental interface of ruminants is not always uniform across a single cross-section of a site of placentation which allows different conclusions to be made depending on the part of the utero-placental interface being examined.
| S-EPMC10895283 | biostudies-literature
Project description:Traditional trajectory learning methods based on Imitation Learning (IL) only learn the existing trajectory knowledge from human demonstration. In this way, it can not adapt the trajectory knowledge to the task environment by interacting with the environment and fine-tuning the policy. To address this problem, a global trajectory learning method which combinines IL with Reinforcement Learning (RL) to adapt the knowledge policy to the environment is proposed. In this paper, IL is proposed to acquire basic trajectory skills, and then learns the agent will explore and exploit more policy which is applicable to the current environment by RL. The basic trajectory skills include the knowledge policy and the time stage information in the whole task space to help learn the time series of the trajectory, and are used to guide the subsequent RL process. Notably, neural networks are not used to model the action policy and the Q value of RL during the RL process. Instead, they are sampled and updated in the whole task space and then transferred to the networks after the RL process through Behavior Cloning (BC) to get continuous and smooth global trajectory policy. The feasibility and the effectiveness of the method was validated in a custom Gym environment of a flower drawing task. And then, we executed the learned policy in the real-world robot drawing experiment.
| S-EPMC10978794 | biostudies-literature
Project description:Summary In drug development, a major reason for attrition is the lack of understanding of cellular mechanisms governing drug toxicity. The black-box nature of conventional classification models has limited their utility in identifying toxicity pathways. Here we developed DTox (deep learning for toxicology), an interpretation framework for knowledge-guided neural networks, which can predict compound response to toxicity assays and infer toxicity pathways of individual compounds. We demonstrate that DTox can achieve the same level of predictive performance as conventional models with a significant improvement in interpretability. Using DTox, we were able to rediscover mechanisms of transcription activation by three nuclear receptors, recapitulate cellular activities induced by aromatase inhibitors and pregnane X receptor (PXR) agonists, and differentiate distinctive mechanisms leading to HepG2 cytotoxicity. Virtual screening by DTox revealed that compounds with predicted cytotoxicity are at higher risk for clinical hepatic phenotypes. In summary, DTox provides a framework for deciphering cellular mechanisms of toxicity in silico. Graphical abstract Highlights • DTox is a deep learning model for toxicity prediction with broad applicability• It provides an interpretation framework to infer toxicity pathways of compounds• We improve the interpretability of toxicity prediction without sacrificing accuracy• DTox’s interpretation framework deciphers cellular mechanisms of toxicity in silico The bigger picture In drug development, a major reason for attrition is the lack of understanding of cellular mechanisms governing drug toxicity. It is challenging to explain the toxicity outcomes of newly developed compounds with limited prior knowledge. To address the challenge, we present DTox (Deep learning for Toxicology), a deep learning model incorporated with extensive knowledge from pathway ontology. DTox is a highly efficient learning model with good predictive performance. It is applicable to all compounds because it requires only chemical structure as model input. More importantly, the knowledge-guided structure of DTox enables us to identify network paths connecting query compounds to toxicity outcomes via target proteins, functional pathways, and general biological processes. Such paths can be viewed as mechanistic interpretation of toxicity and facilitate experimental investigation. We employ existing experimental datasets to validate the mechanistic interpretation by DTox and demonstrate its biological significance. Toxicity assessment is a critical step in drug development. To overcome the black-box nature of conventional classification models, Hao et al. propose an interpretable model named DTox (deep learning for toxicology) for predicting compound response to toxicity assays and inferring toxicity pathways of individual compounds. Validation studies using experimental datasets demonstrate the effectiveness of DTox in rediscovering known mechanisms, differentiating distinctive mechanisms, and recapitulating cellular activities leading to toxicity. DTox will benefit mechanistic studies in toxicology by generating testable hypotheses for further investigation.
| S-EPMC9481960 | biostudies-literature
Project description:Monitoring the health of a pregnancy is of utmost importance to both the fetus and the mother. The diagnosis of pregnancy complications typically occurs after the manifestation of symptoms, and limited preventative measures or effective treatments are available. Traditionally, pregnancy health is evaluated by analyzing maternal serum hormone levels, genetic testing, ultrasonographic imaging, and monitoring maternal symptoms. However, researchers have reported a difference in extracellular vesicle (EV) quantity and cargo between healthy and at-risk pregnancies. Thus, placental EVs (PEVs) may help to understand normal and aberrant placental development, monitor pregnancy health in terms of developing placental pathologies, and assess the impact of environmental influences, such as infection, on pregnancy. The diagnostic potential of PEVs could allow for earlier detection of pregnancy complications via noninvasive sampling and frequent monitoring. Understanding how PEVs serve as a means of communication with maternal cells and recognizing their potential utility as a readout of placental health have sparked a growing interest in basic and translational research. However, to date, PEV research with animal models lags behind human studies. The strength of animal pregnancy models is that they can be used to assess placental pathologies in conjunction with isolation of PEVs from fluid samples at different time points throughout gestation. Assessing PEV cargo in animals within normal and complicated pregnancies will accelerate the translation of PEV analysis into the clinic for potential use in prognostics. We propose that appropriate animal models of human pregnancy complications must be established in the PEV field.
| S-EPMC7786267 | biostudies-literature
Project description:Accurate and cost-effective quantification of the carbon cycle for agroecosystems at decision-relevant scales is critical to mitigating climate change and ensuring sustainable food production. However, conventional process-based or data-driven modeling approaches alone have large prediction uncertainties due to the complex biogeochemical processes to model and the lack of observations to constrain many key state and flux variables. Here we propose a Knowledge-Guided Machine Learning (KGML) framework that addresses the above challenges by integrating knowledge embedded in a process-based model, high-resolution remote sensing observations, and machine learning (ML) techniques. Using the U.S. Corn Belt as a testbed, we demonstrate that KGML can outperform conventional process-based and black-box ML models in quantifying carbon cycle dynamics. Our high-resolution approach quantitatively reveals 86% more spatial detail of soil organic carbon changes than conventional coarse-resolution approaches. Moreover, we outline a protocol for improving KGML via various paths, which can be generalized to develop hybrid models to better predict complex earth system dynamics.
| S-EPMC10774286 | biostudies-literature
Project description:Learning effective molecular feature representation to facilitate molecular property prediction is of great significance for drug discovery. Recently, there has been a surge of interest in pre-training graph neural networks (GNNs) via self-supervised learning techniques to overcome the challenge of data scarcity in molecular property prediction. However, current self-supervised learning-based methods suffer from two main obstacles: the lack of a well-defined self-supervised learning strategy and the limited capacity of GNNs. Here, we propose Knowledge-guided Pre-training of Graph Transformer (KPGT), a self-supervised learning framework to alleviate the aforementioned issues and provide generalizable and robust molecular representations. The KPGT framework integrates a graph transformer specifically designed for molecular graphs and a knowledge-guided pre-training strategy, to fully capture both structural and semantic knowledge of molecules. Through extensive computational tests on 63 datasets, KPGT exhibits superior performance in predicting molecular properties across various domains. Moreover, the practical applicability of KPGT in drug discovery has been validated by identifying potential inhibitors of two antitumor targets: hematopoietic progenitor kinase 1 (HPK1) and fibroblast growth factor receptor 1 (FGFR1). Overall, KPGT can provide a powerful and useful tool for advancing the artificial intelligence (AI)-aided drug discovery process.
| S-EPMC10663446 | biostudies-literature
Project description:The hemochorial placentation site is characterized by a dynamic interplay between trophoblast cells and maternal cells. These cells cooperate to establish an interface required for nutrient delivery to promote fetal growth. In the human, trophoblast cells penetrate deep into the uterus. This is not a consistent feature of hemochorial placentation and has hindered the establishment of suitable animal models. The rat represents an intriguing model for investigating hemochorial placentation with deep trophoblast cell invasion. In this study, we used single-cell RNA sequencing to characterize the transcriptome of the invasive trophoblast cell lineage, as well as other cell populations within the rat uterine-placental interface during early (gestation day [gd] 15.5) and late (gd 19.5) stages of intrauterine trophoblast cell invasion. We identified a robust set of transcripts that define invasive trophoblast cells, as well as transcripts that distinguished endothelial, smooth muscle, natural killer, and macrophage cells. Invasive trophoblast, immune, and endothelial cell populations exhibited distinct spatial relationships within the uterine-placental interface. Furthermore, the maturation stage of invasive trophoblast cell development could be determined by assessing gestation stage-dependent changes in transcript expression. Finally, and most importantly, expression of a prominent subset of rat invasive trophoblast cell transcripts is conserved in the invasive extravillous trophoblast cell lineage of the human placenta. These findings provide foundational data to identify and interrogate key conserved regulatory mechanisms essential for the development and function of an important compartment within the hemochorial placentation site that is essential for a healthy pregnancy.
| S-EPMC9565169 | biostudies-literature
Project description:ObjectivesTo automatically construct a drug indication taxonomy from drug labels using generative Artificial Intelligence (AI) represented by the Large Language Model (LLM) GPT-4 and real-world evidence (RWE).Materials and methodsWe extracted indication terms from 46 421 free-text drug labels using GPT-4, iteratively and recursively generated indication concepts and inferred indication concept-to-concept and concept-to-term subsumption relations by integrating GPT-4 with RWE, and created a drug indication taxonomy. Quantitative and qualitative evaluations involving domain experts were performed for cardiovascular (CVD), Endocrine, and Genitourinary system diseases.Results2909 drug indication terms were extracted and assigned into 24 high-level indication categories (ie, initially generated concepts), each of which was expanded into a sub-taxonomy. For example, the CVD sub-taxonomy contains 242 concepts, spanning a depth of 11, with 170 being leaf nodes. It collectively covers a total of 234 indication terms associated with 189 distinct drugs. The accuracies of GPT-4 on determining the drug indication hierarchy exceeded 0.7 with "good to very good" inter-rater reliability. However, the accuracies of the concept-to-term subsumption relation checking varied greatly, with "fair to moderate" reliability.Discussion and conclusionWe successfully used generative AI and RWE to create a taxonomy, with drug indications adequately consistent with domain expert expectations. We show that LLMs are good at deriving their own concept hierarchies but still fall short in determining the subsumption relations between concepts and terms in unregulated language from free-text drug labels, which is the same hard task for human experts.
| S-EPMC11339527 | biostudies-literature
Project description:The placenta, a unique organ that only develops during pregnancy, is essential for nutrient, oxygen, and waste exchange between offspring and mother. Yet, despite its importance, the placenta remains one of the least understood organs and knowledge of early placental formation is particularly limited. Abnormalities in placental development result in placental dysfunction or insufficiency whereby normal placental physiology is impaired. Placental dysfunction is a frequent source of pregnancy loss in livestock, inflicting serious economic impact to producers. Though the underlying causes of placental dysfunction are not well-characterized, initiation of disease is thought to occur during establishment of functional fetal and placental circulation. A comprehensive understanding of the mechanisms controlling placental growth and vascularization is necessary to improve reproductive success in livestock. We propose chemokine C-X-C motif ligand 12 (CXCL12) signaling through its receptor CXCR4 functions as a chief coordinator of vascularization through direct actions on fetal trophoblast and maternal endometrial and immune cells. To investigate CXCL12-CXCR4 signaling on uteroplacental vascular remodeling at the fetal-maternal interface, we utilized a CXCR4 antagonist (AMD3100). On day 12 post-breeding in sheep, osmotic pumps were surgically installed and delivered either AMD3100 or saline into the uterine lumen ipsilateral to the corpus luteum for 14 days. On day 35 of ovine pregnancy, fetal/placental and endometrial tissues were collected, snap-frozen in liquid nitrogen, and uterine horn cross sections were preserved for immunofluorescent analysis. Suppressing CXCL12-CXCR4 at the fetal-maternal interface during initial placental vascularization resulted in diminished abundance of select angiogenic factors in fetal and maternal placenta on day 35. Compared to control, less vascular endothelial growth factor (VEGF) and VEFG receptor 2 (KDR) were observed in endometrium when CXCL12-CXCR4 was diminished. Less VEGF was also evident in fetal placenta (cotyledons) in ewes receiving AMD3100 infusion compared to control. Suppressing CXCL12-CXCR4 at the fetal-maternal interface also resulted in greater autophagy induction in fetal and maternal placenta compared to control, suggestive of CXCL12-CXCR4 impacting cell survival. CXCL12-CXCR4 signaling may govern placental homeostasis by serving as a critical upstream mediator of vascularization and cell viability, thereby ensuring appropriate placental development.
| S-EPMC8415452 | biostudies-literature
Project description:The hemochorial placentation site is characterized by a dynamic interplay between trophoblast cells and maternal cells. These cells cooperate to establish an interface required for nutrient delivery to promote fetal growth. In the human, trophoblast cells penetrate deep into the uterus. This is not a consistent feature of hemochorial placentation and has hindered the establishment of suitable animal models. The rat represents an intriguing model for investigating hemochorial placentation with deep trophoblast cell invasion. In this study, we used single cell RNA sequencing to characterize the transcriptome of the invasive trophoblast cell lineage, as well as other cell populations within the rat uterine-placental interface during early (gestation day, gd, 15.5) and late (gd 19.5) stages of intrauterine trophoblast cell invasion. We identified a robust set of transcripts that define invasive trophoblast cells, as well as transcripts that distinguished endothelial, smooth muscle, natural killer, and macrophage cells. Invasive trophoblast, immune, and endothelial cell populations exhibited distinct spatial relationships within the uterine-placental interface. Furthermore, the maturation stage of invasive trophoblast cell development could be determined by assessing gestation-stage dependent changes in transcript expression. Finally, and most importantly, expression of a prominent subset of rat invasive trophoblast cell transcripts is conserved in the invasive extravillous trophoblast cell lineage of the human placenta. These findings provide foundational data to identify and interrogate key conserved regulatory mechanisms essential for development and function of an important compartment within the hemochorial placentation site that is essential for a healthy pregnancy.
2022-09-20 | GSE206086 | GEO