Search for: All records

Causal inference from observational data has attracted considerable attention among researchers. One main obstacle is the handling of confounders. As direct measurement of confounders may not be feasible, recent methods seek to address the confounding bias via proxy variables, i.e., covariates postulated to be conducive to the inference of latent confounders. However, the selected proxies may scramble both confounders and post-treatment variables in practice, which risks biasing the estimation by controlling for variables affected by the treatment. In this paper, we systematically investigate the bias due to latent post-treatment variables, i.e., latent post-treatment bias, in causal effect estimation. Specifically, we first derive the bias when selected proxies scramble both latent confounders and post-treatment variables, which we demonstrate can be arbitrarily bad. We then propose a Confounder-identifiable VAE (CiVAE) to address the bias. Based on a mild assumption that the prior of latent variables that generate the proxy belongs to a general exponential family with at least one invertible sufficient statistic in the factorized part, CiVAE individually identifies latent confounders and latent post-treatment variables up to bijective transformations. We then prove that with individual identification, the intractable disentanglement problem of latent confounders and post-treatment variables can be transformed into a tractable independence test problem despite arbitrary dependence may exist among them. Finally, we prove that the true causal effects can be unbiasedly estimated with transformed confounders inferred by CiVAE. Experiments on both simulated and real-world datasets demonstrate significantly improved robustness of CiVAE. 

Job marketplace is a heterogeneous graph composed of interactions among members (job-seekers), companies, and jobs. Understanding and modeling job marketplace can benefit both job seekers and employers, ultimately contributing to the greater good of the society. However, existing graph neural network (GNN)-based methods have shallow understandings of the associated textual features and heterogeneous relations. To address the above challenges, we propose PLM4Job, a job marketplace foundation model that tightly couples pretrained language models (PLM) with job market graph, aiming to fully utilize the pretrained knowledge and reasoning ability to model member/job textual features as well as various member-job relations simultaneously. In the pretraining phase, we propose a heterogeneous ego-graph-based prompting strategy to model and aggregate member/job textual features based on the topological structure around the target member/job node, where entity type embeddings and graph positional embeddings are introduced accordingly to model different entities and their heterogeneous relations. Meanwhile, a proximity-aware attention alignment strategy is designed to dynamically adjust the attention of the PLM on ego-graph node tokens in the prompt, such that the attention can be better aligned with job marketplace semantics. Extensive experiments at LinkedIn demonstrate the effectiveness of PLM4Job. 

Abstract Research on elemental distribution in plants is crucial for understanding nutrient uptake, environmental adaptation and optimizing agricultural practices for sustainable food production. Plant trichomes, with their self-contained structures and easy accessibility, offer a robust model system for investigating elemental repartitioning. Transport proteins, such as the four functional cation exchangers (CAXs) in Arabidopsis, are low-affinity, high-capacity transporters primarily located on the vacuole. Mutants in these transporters have been partially characterized, one of the phenotypes of the CAX1 mutant being altered with tolerance to low-oxygen conditions. A simple visual screen demonstrated trichome density and morphology in cax1, and quadruple CAX (cax1-4: qKO) mutants remained unaltered. Here, we used synchrotron X-ray fluorescence (SXRF) to show that trichomes in CAX-deficient lines accumulated high levels of chlorine, potassium, calcium and manganese. Proteomic analysis on isolated Arabidopsis trichomes showed changes in protein abundance in response to changes in element accumulation. The CAX mutants showed an increased abundance of plasma membrane ATPase and vacuolar H-pumping proteins, and proteins associated with water movement and endocytosis, while also showing changes in proteins associated with the regulation of plasmodesmata. These findings advance our understanding of the integration of CAX transport with elemental homeostasis within trichomes and shed light on how plants modulate protein abundance under conditions of altered elemental levels. 

Recently, there has been growing interest in developing the next-generation recommender systems (RSs) based on pretrained large language models (LLMs). However, the semantic gap between natural language and recommendation tasks is still not well addressed, leading to multiple issues such as spuriously correlated user/item descriptors, ineffective language modeling on user/item data, inefficient recommendations via auto-regression, etc. In this paper, we propose CLLM4Rec, the first generative RS that tightly integrates the LLM paradigm and ID paradigm of RSs, aiming to address the above challenges simultaneously. We first extend the vocabulary of pretrained LLMs with user/item ID tokens to faithfully model user/item collaborative and content semantics. Accordingly, a novel soft+hard prompting strategy is proposed to effectively learn user/item collaborative/content token embeddings via language modeling on RS-specific corpora, where each document is split into a prompt consisting of heterogeneous soft (user/item) tokens and hard (vocab) tokens and a main text consisting of homogeneous item tokens or vocab tokens to facilitate stable and effective language modeling. In addition, a novel mutual regularization strategy is introduced to encourage CLLM4Rec to capture recommendation-related information from noisy user/item content. Finally, we propose a novel recommendation-oriented finetuning strategy for CLLM4Rec, where an item prediction head with multinomial likelihood is added to the pretrained CLLM4Rec backbone to predict hold-out items based on soft+hard prompts established from masked user-item interaction history, where recommendations of multiple items can be generated efficiently without hallucination. 

Recommender systems (RSs) have become an indispensable part of online platforms. With the growing concerns of algorithmic fairness, RSs are not only expected to deliver high-quality personalized content, but are also demanded not to discriminate against users based on their demographic information. However, existing RSs could capture undesirable correlations between sensitive features and observed user behaviors, leading to biased recommendations. Most fair RSs tackle this problem by completely blocking the influences of sensitive features on recommendations. But since sensitive features may also affect user interests in a fair manner (e.g., race on culture-based preferences), indiscriminately eliminating all the influences of sensitive features inevitably degenerate the recommendations quality and necessary diversities. To address this challenge, we propose a path-specific fair RS (PSF-RS) for recommendations. Specifically, we summarize all fair and unfair correlations between sensitive features and observed ratings into two latent proxy mediators, where the concept of path-specific bias (PS-Bias) is defined based on path-specific counterfactual inference. Inspired by Pearl's minimal change principle, we address the PS-Bias by minimally transforming the biased factual world into a hypothetically fair world, where a fair RS model can be learned accordingly by solving a constrained optimization problem. For the technical part, we propose a feasible implementation of PSF-RS, i.e., PSF-VAE, with weakly-supervised variational inference, which robustly infers the latent mediators such that unfairness can be mitigated while necessary recommendation diversities can be maximally preserved simultaneously. Experiments conducted on semi-simulated and real-world datasets demonstrate the effectiveness of PSF-RS. 

Abstract Multiple Arabidopsis H⁺/Cation exchangers (CAXs) participate in high‐capacity transport into the vacuole. Previous studies have analysed single and double mutants that marginally reduced transport; however, assessing phenotypes caused by transport loss has proven enigmatic. Here, we generated quadruple mutants (cax1‐4: qKO) that exhibited growth inhibition, an 85% reduction in tonoplast‐localised H⁺/Ca transport, and enhanced tolerance to anoxic conditions compared to CAX1 mutants. Leveraging inductively coupled plasma mass spectrometry (ICP‐MS) and synchrotron X‐ray fluorescence (SXRF), we demonstrate CAX transporters work together to regulate leaf elemental content: ICP‐MS analysis showed that the elemental concentrations in leaves strongly correlated with the number of CAX mutations; SXRF imaging showed changes in element partitioning not present in single CAX mutants and qKO had a 40% reduction in calcium (Ca) abundance. Reduced endogenous Ca may promote anoxia tolerance; wild‐type plants grown in Ca‐limited conditions were anoxia tolerant. Sequential reduction of CAXs increased mRNA expression and protein abundance changes associated with reactive oxygen species and stress signalling pathways. Multiple CAXs participate in postanoxia recovery as their concerted removal heightened changes in postanoxia Ca signalling. This work showcases the integrated and diverse function of H⁺/Cation transporters and demonstrates the ability to improve anoxia tolerance through diminishing endogenous Ca levels. 

Several COVID-19 vaccines have been on the market since early 2021 and may vary in their effectiveness and safety. This study characterizes hesitancy about accepting COVID-19 vaccines among parents in Shanghai, China, and identifies how sensitive they are to changes in vaccine safety and effectiveness profiles. Schools in each township of Minhang District, Shanghai, were sampled, and parents in the WeChat group of each school were asked to participate in this cross-sectional Internet-based survey. Parents responded to questions about hesitancy and were given information about five different COVID-19 vaccine candidates, the effectiveness of which varied between 50 and 95% and which had a risk of fever as a side effect between 5 and 20%. Overall, 3673 parents responded to the survey. Almost 90% would accept a vaccine for themselves (89.7%), for their child (87.5%) or for an elderly parent (88.5%) with the most ideal attributes (95% effectiveness with 5% risk of fever). But with the least ideal attributes (50% effectiveness and a 20% risk of fever) these numbers dropped to 33.5%, 31.3%, and 31.8%, respectively. Vaccine hesitancy, age at first child’s birth, and relative income were all significantly related to sensitivity to vaccine safety and effectiveness. Parents showed a substantial shift in attitudes towards a vaccine based on its safety and effectiveness profile. These findings indicate that COVID-19 vaccine acceptance may be heavily influenced by how effective the vaccine actually is and could be impeded or enhanced based on vaccines already on the market. 

Abstract A plant’s oxygen supply can vary from normal (normoxia) to total depletion (anoxia). Tolerance to anoxia is relevant to wetland species, rice (Oryza sativa) cultivation, and submergence tolerance of crops. Decoding and transmitting calcium (Ca) signals may be an important component to anoxia tolerance; however, the contribution of intracellular Ca transporters to this process is poorly understood. Four functional cation/proton exchangers (CAX1–4) in Arabidopsis (Arabidopsis thaliana) help regulate Ca homeostasis around the vacuole. Our results demonstrate that cax1 mutants are more tolerant to both anoxic conditions and submergence. Using phenotypic measurements, RNA-sequencing, and proteomic approaches, we identified cax1-mediated anoxia changes that phenocopy changes present in anoxia-tolerant crops: altered metabolic processes, diminished reactive oxygen species production post anoxia, and altered hormone signaling. Comparing wild-type and cax1 expressing genetically encoded Ca indicators demonstrated altered cytosolic Ca signals in cax1 during reoxygenation. Anoxia-induced Ca signals around the plant vacuole are involved in the control of numerous signaling events related to adaptation to low oxygen stress. This work suggests that cax1 anoxia response pathway could be engineered to circumvent the adverse effects of flooding that impair production agriculture.