Search for: All records

Graph Neural Networks (GNNs) have been increasingly deployed in a plethora of applications. However, the graph data used for training may contain sensitive personal information of the involved individuals. Once trained, GNNs typically encode such information in their learnable parameters. As a consequence, privacy leakage may happen when the trained GNNs are deployed and exposed to potential attackers. Facing such a threat, machine unlearning for GNNs has become an emerging technique that aims to remove certain personal information from a trained GNN. Among these techniques, certified unlearning stands out, as it provides a solid theoretical guarantee of the information removal effectiveness. Nevertheless, most of the existing certified unlearning methods for GNNs are only designed to handle node and edge unlearning requests. Meanwhile, these approaches are usually tailored for either a specific design of GNN or a specially designed training objective. These disadvantages significantly jeopardize their flexibility. In this paper, we propose a principled framework named IDEA to achieve flexible and certified unlearning for GNNs. Specifically, we first instantiate four types of unlearning requests on graphs, and then we propose an approximation approach to flexibly handle these unlearning requests over diverse GNNs. We further provide theoretical guarantee of the effectiveness for the proposed approach as a certification. Different from existing alternatives, IDEA is not designed for any specific GNNs or optimization objectives to perform certified unlearning, and thus can be easily generalized. Extensive experiments on real-world datasets demonstrate the superiority of IDEA in multiple key perspectives. 

To date, the presence of pharmaceuticals has been extensively documented across a wide range of aquatic systems and biota. Further, substantial progress has been made in transitioning from laboratory assessments of pharmaceutical fate and effects in fish to in situ assessments of exposure and effects; however, certain research areas remain understudied. Among these is investigation of differential accumulation across multiple internal tissues in wild marine fish beyond the species commonly sampled in laboratory and freshwater field settings. This study examined the presence of pharmaceuticals across four tissues (plasma, muscle, brain, and liver) in a wild marine fish, bonefish (Albula vulpes), throughout coastal South Florida, USA. Differential accumulation across tissues was assessed for the number and concentration, identity, and composition of accumulated pharmaceuticals by sampling 25 bonefish and analyzing them for 91 pharmaceuticals. The concentration of pharmaceuticals was highest in plasma > liver > brain > muscle, while the number of pharmaceuticals was highest in liver > brain > plasma > muscle. The identity of detected pharmaceuticals was tissue specific, and there was an inverse relationship between the number of detections for each pharmaceutical and its log Kow. The composition of pharmaceuticals was tissue specific for both pharmaceutical presence/absence and concentration. Across all tissues, the greatest similarity was between brain and liver, which were more similar to plasma than to muscle, and muscle was the most distinct tissue. For tissue compositional variability, muscle was the most diverse in accumulated pharmaceuticals, while plasma, brain, and liver were similarly variable. With the highest concentrations in plasma and highest number in liver, and documented variability in accumulated pharmaceuticals across tissues, our results highlight the importance of tissue selection when surveying exposure in wild fish, suggesting that multi-tissue analysis would allow for a more comprehensive assessment of exposure diversity and risk of adverse effects. 

The commonly used analytical tools for metabolomics cannot directly probe metabolic activities or distinguish metabolite differences between cells and suborgans in multicellular organisms. These issues can be addressed byin-vivoisotope labeling and mass spectrometry imaging (MSI), respectively, but the combination of the two, a newly emerging technology we call MSIi, has been rarely applied to plant systems. In this study, we explored MSIiofArabidopsis thalianawith D₂O labeling to study and visualize D-labeling in three classes of lipids: arabidopsides, chloroplast lipids, and epicuticular wax. Similar to other stress responses, D₂O-induced stress increased arabidopsides in an hour, but it was relatively minor for matured plants and reverted to the normal level in a few hours. The D-labeling isotopologue patterns of arabidopsides matched with those of galactolipid precursors, supporting the currently accepted biosynthesis mechanism. Matrix-assisted laser desorption/ionization (MALDI)-MSI was used to visualize the spatiotemporal distribution of deuterated chloroplast lipids, pheophytina, MGDGs, and DGDGs, after growing day-after-sowing (DAS) 28 plants in D₂O condition for 3–12 days. There was a gradual change of deuteration amount along the leaf tissues and with a longer labeling time, which was attributed to slow respiration leading to low D₂O concentration in the tissues. Finally, deuterium incorporation in epicuticular wax was visualized on the surfaces of the stem and flower. The conversion efficiency of newly synthesized C30 aldehyde to C29 ketone was very low in the lower stem but very high at the top of the stem near the flower or on the flower carpel. This study successfully demonstrated that MSIican unveil spatiotemporal metabolic activities in various tissues ofA. thaliana.

 

Pharmaceutical uptake involves processes that vary across aquatic systems and biota. However, single studies examining multiple environmental compartments, microhabitats, biota, and exposure pathways in mesoconsumer fish are sparse. We investigated the pharmaceutical burden in bonefish (Albula vulpes), pathways of exposure, and estimated exposure to a human daily dose. To evaluate exposure pathways, the number and composition of pharmaceuticals across compartments and the bioconcentration in prey and bonefish were assessed. To evaluate bioaccumulation, we proposed the use of a field-derived bioaccumulation factor (fBAF), due to variability inherent to natural systems. Exposure to a human daily dose was based on bonefish daily energetic requirements and consumption rates using pharmaceutical concentrations in prey. Pharmaceutical number and concentration were highest in prey, followed by bonefish, water and sediment. Fifteen pharmaceuticals were detected in common among bonefish, prey, and water; all of which bioconcentrated in prey and bonefish, and four bioaccumulated in bonefish. The composition of detected pharmaceuticals was compartment specific, and prey were most similar to bonefish. Bonefish were exposed to a maximum of 1.2 % of a human daily dose via prey consumption. Results highlight the need for multicompartment assessments of exposure and consideration of prey along with water as a pathway of exposure. 

A pentadentate pyridinophane ligand is reported, along with a palladacycle product generatedviaCsp³–H bond activation under mild conditions.

 

The interaction between T-cell receptors (TCRs) and peptides (epitopes) presented by major histocompatibility complex molecules (MHC) is fundamental to the immune response. Accurate prediction of TCR–epitope interactions is crucial for advancing the understanding of various diseases and their prevention and treatment. Existing methods primarily rely on sequence-based approaches, overlooking the inherent topology structure of TCR–epitope interaction networks. In this study, we present $GTE$, a novel heterogeneous Graph neural network model based on inductive learning to capture the topological structure between TCRs and Epitopes. Furthermore, we address the challenge of constructing negative samples within the graph by proposing a dynamic edge update strategy, enhancing model learning with the nonbinding TCR–epitope pairs. Additionally, to overcome data imbalance, we adapt the Deep AUC Maximization strategy to the graph domain. Extensive experiments are conducted on four public datasets to demonstrate the superiority of exploring underlying topological structures in predicting TCR–epitope interactions, illustrating the benefits of delving into complex molecular networks. The implementation code and data are available at https://github.com/uta-smile/GTE.

 

Plant microbiomes that comprise diverse microorganisms, including prokaryotes, eukaryotes and viruses, are the key determinants of plant population dynamics and ecosystem function. Despite their importance, little is known about how species interactions (especially trophic interactions) between microbes from different domains modify the importance of microbiomes for plant hosts and ecosystems. Using the common duckweedLemna minor, we experimentally examined the effects of predation (by bacterivorous protists) and parasitism (by bacteriophages) within microbiomes on plant population size and ecosystem phosphorus removal. Our results revealed that the addition of predators increased plant population size and phosphorus removal, whereas the addition of parasites showed the opposite pattern. The structural equation modelling further pointed out that predation and parasitism affected plant population size and ecosystem function via distinct mechanisms that were both mediated by microbiomes. Our results highlight the importance of understanding microbial trophic interactions for predicting the outcomes and ecosystem impacts of plant–microbiome symbiosis.