Context-dependent Research Articles

The Social Construction of Categorical Data: Mixed Methods Approach to Assessing Data Features in Publicly Available Datasets.

In data-sparse areas such as health care, computer scientists aim to leverage as much available information as possible to increase the accuracy of their machine learning models' outputs. As a standard, categorical data, such as patients' gender, socioeconomic status, or skin color, are used to train models in fusion with other data types, such as medical images and text-based medical information. However, the effects of including categorical data features for model training in such data-scarce areas are underexamined, particularly regarding models intended to serve individuals equitably in a diverse population. This study aimed to explore categorical data's effects on machine learning model outputs, rooted the effects in the data collection and dataset publication processes, and proposed a mixed methods approach to examining datasets' data categories before using them for machine learning training. Against the theoretical background of thesocial construction of categories, we suggest a mixed methods approach to assess categorical data's utility for machine learning model training. As an example, we applied our approach to a Brazilian dermatological dataset (Dermatological and Surgical Assistance Program at the Federal University of Espírito Santo [PAD-UFES] 20). We first present an exploratory, quantitative study that assesses the effects when including or excluding each of the unique categorical data features of the PAD-UFES 20 dataset for training a transformer-based model using a data fusion algorithm. We then pair our quantitative analysis with a qualitative examination of the data categories based on interviews with the dataset authors. Our quantitative study suggests scattered effects of including categorical data for machine learning model training across predictive classes. Our qualitative analysis gives insights into how the categorical data were collected and why they were published, explaining some of the quantitative effects that we observed. Our findings highlight the social constructedness of categorical data in publicly available datasets, meaning that the data in a category heavily depend on both how these categories are defined by the dataset creators and the sociomedico context in which the data are collected. This reveals relevant limitations of using publicly available datasets in contexts different from those of the collection of their data. We caution against using data features of publicly available datasets without reflection on the social construction and context dependency of their categorical data features, particularly in data-sparse areas. We conclude that social scientific, context-dependent analysis of available data features using both quantitative and qualitative methods is helpful in judging the utility of categorical data for the population for which a model is intended.

Tibetan Judicial Event Extraction Based on Deep Word Representation and Hybrid Neural Networks

For the extraction of judicial events for Tibetan, a low-resource language, traditional simple neural network approaches struggle to adequately capture the deep semantics and features of the texts because Tibetan texts are usually lengthy and contain numerous judicial-related entities. To overcome this limitation, this research presents an event extraction model combining deep word representation with hybrid neural networks for the Tibetan judicial domain. The model introduces the Chinese minority pre-trained language model (CINO), which generates dynamic word vector representations, addressing the challenge of modeling the deep semantics inherent in Tibetan texts. During feature extraction, a bidirectional long short-term memory network (BiLSTM) is applied to extract the temporal and contextual dependencies, while a convolutional neural network (CNN) is utilized to capture the local semantic features to construct a comprehensive global semantic representation. Finally, the sequences are decoded through conditional random field (CRF) to generate optimal prediction results, thus achieving the efficient extraction of Tibetan judicial events. The experimental findings indicate that the model outperforms the baselines by achieving F1 scores of 70.47% for trigger detection and 62.99% for argument recognition, with improvements of 16.6% and 16.42%, respectively. These results confirm the effectiveness and superiority of the proposed model.

Language proficiency and cultural intelligence: A meta‐analysis based on conservation of resources theory and the overt‐covert model of culture

AbstractResearch on language proficiency and cultural intelligence (CQ) has developed somewhat independently, resulting in a partial understanding of the personal resources required to facilitate job performance and adaptation in a foreign culture. Our meta‐analysis synthesizes the two constructs by drawing on the overt‐covert model of culture and the conservation of resources (COR) theory. Specifically, we propose that language proficiency and CQ complement each other by overcoming resource losses arising from overt (i.e., explicit language differences) and covert (i.e., implicit rules) cultural differences, respectively. We further propose that the host‐country communication context (i.e., higher vs. lower context dependence) moderates the effects of language proficiency and CQ. Meta‐analytic results based on 355 field samples (N = 128,358) support our hypotheses and show that (1) both language proficiency and CQ are positively related to job performance and cultural adaptation, (2) the effects of language proficiency are accentuated in cultures with lower communication‐context dependence, while (3) the effects of CQ are accentuated in cultures with higher communication‐context dependence. We discuss the implications of these findings for future research and practice.

A quantitative intracellular peptide binding assay reveals recognition determinants and context dependence of short linear motifs.

Transient protein-protein interactions play key roles in controlling dynamic cellular responses. Many examples involve globular protein domains that bind to peptide sequences known as Short Linear Motifs (SLiMs), which are enriched in intrinsically disordered regions of proteins. Here we describe a novel functional assay for measuring SLiM binding, called Systematic Intracellular Motif Binding Analysis (SIMBA). In this method, binding of a foreign globular domain to its cognate SLiM peptide allows yeast cells to proliferate by blocking a growth arrest signal. A high-throughput application of the SIMBA method involving competitive growth and deep sequencing provides rapid quantification of the relative binding strength for thousands of SLiM sequence variants, and a comprehensive interrogation of SLiM sequence features that control their recognition and potency. We show that multiple distinct classes of SLiM-binding domains can be analyzed by this method, and that the relative binding strength of peptides in vivo correlates with their biochemical affinities measured in vitro. Deep mutational scanning provides high-resolution definitions of motif recognition determinants and reveals how sequence variations at non-core positions can modulate binding strength. Furthermore, mutational scanning of multiple parent peptides that bind human tankyrase ARC or YAP WW domains identifies distinct binding modes and uncovers context effects in which the preferred residues at one position depend on residues elsewhere. The findings establish SIMBA as a fast and incisive approach for interrogating SLiM recognition via massively parallel quantification of protein-peptide binding strength in vivo.

Enhancing concealed object detection in active THz security images with adaptation-YOLO

The terahertz (THz) security scanner offers advantages such as non-contact inspection and the ability to detect various types of dangerous goods, playing an important role in preventing terrorist attacks. We aim to accurately and quickly detect concealed objects in THz security images. However, current object detection algorithms face many challenges when applied to THz images. The main reasons for the detection difficulty are that the concealed objects are small, the image resolution is low, and there is back-ground noise. Many methods often ignore the contextual dependency of the objects, hindering the effective capture of the object’s features. To address this task, this paper first proposes an adaptive context-aware attention network (ACAN), which models global contextual association features in both spatial and channel dimensions. By dynamically combining local features and their global relationships, contextual association information can be obtained from the input features, and enhanced attention features can be achieved through feature fusion to enable precise detection of concealed objects. Secondly, we improved the adaptive convolution and developed the dynamic adaptive convolution block (DACB). DACB can adaptively adjust convolution filter parameters and allocate the filters to the corresponding spatial regions, then filter the feature maps to suppress interference information. Finally, we integrated these two components to YOLOv8, resulting in Adaptation-YOLO. Through wide-ranging experiments on the active THz image dataset, the results demonstrate that the suggested method effectively improves the accuracy and efficiency of object detectors.

Cultural Sensitivity in AI Language Learning: Using NLP to Enhance Language Understanding Across Cultural Contexts

The aim of this paper is to demonstrate the integration of cultural awareness into AI language learning systems using Natural Language Processing (NLP) models. Recognising that the traditional methods of language learning tend not to capture cultural variations, the studies focus on how culturally relevant NLP models (GPT-3, BERT, RNNs) promote language acquisition and cross-cultural integration. Data preprocessing, cultural data augmentation and model training was used to encode cultural variables such as greetings, politeness, and contextually relevant expressions into training datasets. Experimental results suggest that culturally sensitive models outperform the generic model on tasks that involve cultural awareness, such as politeness detection, cultural phrase recognition, and tone sensitivity. GPT-3 for example improved the accuracy of politeness detection tasks from 72.4% to 85.7%. These results highlight the need to bring cultural awareness into AI to develop engaging, context-sensitive language learning. It concludes that culturally based NLP models enhance learner capacity for real-world communication and foster global cultural awareness. Future work should focus on expanding culturally diverse datasets and refining AI models to address subtle cultural complexities

Partially observable Markov models inferred using statistical tests reveal context-dependent syllable transitions in Bengalese finch songs.

Generative models have diverse applications, including language processing and birdsong analysis. In this study, we demonstrate how a statistical test, designed to prevent overgeneralization in sequence generation, can be used to infer minimal models for the syllable sequences in Bengalese finch songs. We focus on the partially observable Markov model (POMM), which consists of states and the probabilistic transitions between them. Each state is associated with a specific syllable, with the possibility that multiple states may correspond to the same syllable. This characteristic differentiates the POMM from a standard Markov model, where each syllable is linked to a single state. The presence of multiple states for a syllable suggests that transitions between syllables are influenced by the specific contexts in which these transitions occur. We apply this method to analyze the songs of six adult male Bengalese finches, both before and after they were deafened. Our results indicate that auditory feedback plays a crucial role in shaping the context-dependent syllable transitions characteristic of Bengalese finch songs.Significance statement Generative models are proficient at representing sequences where the order of elements, such as words or birdsong syllables, is context-dependent. In this study, we demonstrate that a probabilistic model, inspired by the neural encoding involved in song production in songbirds, effectively captures the context-dependent transitions of syllables in Bengalese finch songs. Our findings reveal that the absence of auditory input, as observed in deafened Bengalese finches, reduces these context dependencies, indicating that auditory feedback is essential for establishing context-dependent sequencing in their songs. This method can be applied to various behavioral sequences, offering valuable insights into the neural mechanisms underlying the statistical patterns that govern these sequences.

Program Analysis via Multiple Context Free Language Reachability

Context-free language (CFL) reachability is a standard approach in static analyses, where the analysis question (e.g., is there a dataflow from x to y ?) is phrased as a language reachability problem on a graph G wrt a CFL L . However, CFLs lack the expressiveness needed for high analysis precision. On the other hand, common formalisms for context-sensitive languages are too expressive, in the sense that the corresponding reachability problem becomes undecidable. Are there useful context-sensitive language-reachability models for static analysis? In this paper, we introduce Multiple Context-Free Language (MCFL) reachability as an expressive yet tractable model for static program analysis. MCFLs form an infinite hierarchy of mildly context sensitive languages parameterized by a dimension d and a rank r . Larger d and r yield progressively more expressive MCFLs, offering tunable analysis precision. We showcase the utility of MCFL reachability by developing a family of MCFLs that approximate interleaved Dyck reachability, a common but undecidable static analysis problem. Given the increased expressiveness of MCFLs, one natural question pertains to their algorithmic complexity, i.e., how fast can MCFL reachability be computed? We show that the problem takes O ( n 2 d +1 ) time on a graph of n nodes when r =1, and O ( n d ( r +1) ) time when r >1. Moreover, we show that when r =1, even the simpler membership problem has a lower bound of n 2 d based on the Strong Exponential Time Hypothesis, while reachability for d =1 has a lower bound of n 3 based on the combinatorial Boolean Matrix Multiplication Hypothesis. Thus, for r =1, our algorithm is optimal within a factor n for all levels of the hierarchy based on the dimension d (and fully optimal for d =1). We implement our MCFL reachability algorithm and evaluate it by underapproximating interleaved Dyck reachability for a standard taint analysis for Android. When combined with existing overapproximate methods, MCFL reachability discovers all tainted information on 8 out of 11 benchmarks, while it has remarkable coverage (confirming 94.3% of the reachable pairs reported by the overapproximation) on the remaining 3. To our knowledge, this is the first report of high and provable coverage for this challenging benchmark set.

Indirect Effects and Context Dependency in Stream Fish Invasions

ABSTRACTAimInvasion ecology is replete with a body of well‐supported yet contradictory evidence for numerous invasion hypotheses, likely as a result of context dependency. Context dependency in invasion studies can arise in two ways: (1) apparent, when results differ between studies solely due to methodical differences, or (2) mechanistic, when results truly differ due to ecological processes. One form of apparent context dependency occurs when causally linked factors associated with invasion success (hereafter, invasion drivers) either mask or enhance each other's effect on invasion success. Mechanistic context dependency can occur when regional scale processes modify the influence of local scale invasion processes. Together, apparent and mechanistic context dependency likely give rise to conflicting support between invasion hypotheses via confounding effects of causally related invasion drivers and region‐specific invasion processes.Location2339 stream segments in two ecoregions of the United States.MethodsUsing local scale stream fish community data for two distinct ecoregions, we constructed identical path models to estimate the direct and indirect effects of invasion drivers on nonnative richness. We chose one variable to index invasion drivers from each of the following categories: propagule pressure, natural abiotic, anthropogenic abiotic and biotic factors.ResultsWe found evidence of apparent context dependency through the presence of indirect effects, in which the effects of propagule pressure and biotic factors on nonnative richness were modulated by abiotic factors. The indirect effects of invasion drivers differed between both regions, providing evidence of mechanistic context dependency.Main ConclusionsApparent and mechanistic context dependency can lead to conflicting evidence between studies of invasion hypotheses. Accounting for indirect effects of invasion drivers is important in gaining a more general understanding of the invasion process. Furthermore, because indirect effects varied regionally, it is important to understand the large‐scale processes that contextualise local invasion processes.

Multilingual pretrained based multi-feature fusion model for English text classification

Deep learning methods have been widely applied to English text classification tasks in recent years, achieving strong performance. However, current methods face two significant challenges: (1) they struggle to effectively capture long-range contextual structure information within text sequences, and (2) they do not adequately integrate linguistic knowledge into representations for enhancing the performance of classifiers. To this end, a novel multilingual pre-training based multi-feature fusion method is proposed for English text classification (MFFMP-ETC). Specifically, MFFMP-ETC consists of the multilingual feature extraction, the multilevel structure learning, and the multi-view representation fusion. MFFMP-ETC utilizes the Multilingual BERT as deep semantic extractor to introduce language information into representation learning, which significantly endows text representations with robustness. Then, MFFMP-ETC integrates Bi-LSTM and TextCNN into multilingual pre-training architecture to capture global and local structure information of English texts, via modelling bidirectional contextual semantic dependencies and multi-granularity local semantic dependencies. Meanwhile, MFFMP-ETC devises the multi-view representation fusion within the invariant semantic learning of representations to aggregate consistent and complementary information among views. MFFMP-ETC synergistically integrates Multilingual BERT?s deep semantic features, Bi-LSTM?s bidirectional context processing, and TextCNN local feature extraction, offering a more comprehensive and effective solution for capturing long-distance dependencies and nuanced contextual information in text classification. Finally, results on three datasets show MFFMP-ETC conducts a new baseline in terms of accuracy, sensitivity, and precision, verifying progressiveness and effectiveness of MFFMP-ETC in the text classification.

Enhancing Handwritten Digit Recognition using Auxiliary Classifier Generative Adversarial Networks and Self-attention Mechanism

This research study investigates the integration of the self-attention mechanism and the Auxiliary Classifier Generative Adversarial Networks (ACGAN)to improve handwritten digital recognition using the MNIST data set. Although progression has been made in the generative Adversarial Networks (GANs) used for image synthesis, it is still tested to attain top-notch, context-accurate photo generation, specifically under various information problems. This study fixes these spaces by incorporating self-attention with ACGANs, improving the integrity and labelling accuracy of the produced images. This approach entails changing the ACGAN framework to incorporate the self-focus module, so far better identify the context dependencies in the photo. This assimilation advertises much more detailed and accurate numerical depiction, which is especially helpful when converting data sets with dark histories right into more clear, segmented white background pictures, which enhances the differentiation between data classes. The outcomes reveal that the recognition accuracy and processing speed have been significantly boosted, validating the adaptability and usefulness of the design in various procedure circumstances. The research results show that this method can establish brand-new standards for producing high-grade electronic photos in numerous applications and show considerable progress in the field of image recognition technology.

Similar Conditions With Opposite Effects: Predation-Risk Effects on Prey Abundance Are Highly Contingent.

Experiments have shown that predation-risk effects on prey fitness can be highly contingent on environmental conditions, suggesting a potential difficulty in generalizing risk effects on prey abundance in natural settings. Rather than study the influence of a particular controlled factor, we examine the problem with a novel approach. We examined the influence of risk effects in multiple experiments performed under similar study conditions. Any differences in the experiments would typically be deemed incidental, that is, they would not be given attention in methodology, nor be presented as factors affecting results or inferences. Therefore, any differences in the magnitude and direction of risk effects among experiments would indicate that risk effects on prey population abundance are strongly influenced by context in natural communities. The multiple experiments were conducted under similar conditions, objectives, measurables and implementation, and captured much of the complexity of natural systems (e.g., they were performed with diverse prey assemblages (≥ 11 taxa) over multiple prey generations). Our results highlight the potentially profound context dependence of risk effects: risk effects on the density of some zooplankton species varied between a significant negative effect in one experiment to a significant positive effect in another, whereas other species showed significant negative or positive effects in one experiment and no effect in another. We review mechanisms that could underlie risk effects having opposite effects on the same prey. Our findings illustrate that risk effects observed in one study may not hold, even for the same species in the same system.

How does forest fertilization influence tree productivity of boreal forests? An analysis of data from commercial forestry across Sweden.

Forest fertilization is a forest management practice that is often claimed to increase productivity in boreal forests. Although regarded as an efficient way to increase profitability, it is also costly, and associated with risks such as biodiversity loss and nitrogen leaching from the soil. To be both cost-efficient and sustainable, potential enhanced productivity due to fertilization should be balanced against the adverse environmental impact. One effective strategy is to limit fertilization to sites where it can most significantly increase tree productivity, while avoiding application in less suitable areas. However, the current understanding of the specific conditions under which forest fertilization optimally stimulates tree growth is limited. To clarify this, we analysed standing tree volume from 32,498 recently harvested fertilized and unfertilized stands from Sweden's largest forest owner. We applied generalized additive models to quantify the effect of fertilization on standing tree volume at harvest and how the fertilization effect depended on dominant tree species, stand characteristics (site index, stem density), climatic conditions (temperature sum), and management (thinning, stand age at harvest). We found that the effect of operational fertilization was highly context-dependent. In pine-dominated stands, fertilization failed to increase tree volumes in cold climates and low-productive stands. In spruce-dominated stands, fertilization did not result in increased tree volumes in low-productive and high-productive stands. For a more sustainable and cost-efficient application of this practice we suggest that the context dependency of the efficiency of fertilization is given more attention. Hence, we recommend to refrain from fertilizing pine-dominated stands situated on low-productive land or in regions with cold climates, such as those found in northern Sweden.

Automatic scent creation by cheminformatics method

The sense of smell is fundamental for various aspects of human existence including the flavor perception, environmental awareness, and emotional impact. However, unlike other senses, it has not been digitized. Its digitalization faces challenges such as the lack of reliable odor sensing technology or the precise scent delivery through olfactory displays. Its subjective nature and context dependence add complexity to the process. Moreover, the method of converting odors to digital information remains unclear. This work focuses on one of the most challenging aspects of digital olfaction: automatic scent creation. We propose a method that automatically creates a desired odor profile with the addition of one specific odor descriptor. It is based on a deep neural network that predicts odor descriptors from the multidimensional sensing data, such as mass spectra and an odor reproduction technique using odor components. The results demonstrate that the proposed method can successfully create a scent with the desired odor profile and that its performance depends on the accuracy of the underlying odor predicting method. This opens up the possibility of automatic scent creation, allowing for the presentation of scents with specific odor profiles with an olfactory display.

MultiPhys: Heterogeneous Fusion of Mamba and Transformer for Video-Based Multi-Task Physiological Measurement.

Due to its non-contact characteristics, remote photoplethysmography (rPPG) has attracted widespread attention in recent years, and has been widely applied for remote physiological measurements. However, most of the existing rPPG models are unable to estimate multiple physiological signals simultaneously, and the performance of the limited available multi-task models is also restricted due to their single-model architectures. To address the above problems, this study proposes MultiPhys, adopting a heterogeneous network fusion approach for its development. Specifically, a Convolutional Neural Network (CNN) is used to quickly extract local features in the early stage, a transformer captures global context and long-distance dependencies, and Mamba is used to compensate for the transformer's deficiencies, reducing the computational complexity and improving the accuracy of the model. Additionally, a gate is utilized for feature selection, which classifies the features of different physiological indicators. Finally, physiological indicators are estimated after passing features to each task-related head. Experiments on three datasets show that MultiPhys has superior performance in handling multiple tasks. The results of cross-dataset and hyper-parameter sensitivity tests also verify its generalization ability and robustness, respectively. MultiPhys can be considered as an effective solution for remote physiological estimation, thus promoting the development of this field.

What ChatGPT Has to Say About Its Topological Structure: The Anyon Hypothesis

Large language models (LLMs) achieve remarkable predictive capabilities but remain opaque in their internal reasoning, creating a pressing need for more interpretable artificial intelligence. Here, we propose bridging this explanatory gap by drawing on concepts from topological quantum computing (TQC), specifically the anyonic frameworks arising from SU(2)k theories. Anyons interpolate between fermions and bosons, offering a mathematical language that may illuminate the latent structure and decision-making processes within LLMs. By examining how these topological constructs relate to token interactions and contextual dependencies in neural architectures, we aim to provide a fresh perspective on how meaning and coherence emerge. After eliciting insights from ChatGPT and exploring low-level cases of SU(2)k models, we argue that the machinery of modular tensor categories and topological phases could inform more transparent, stable, and robust AI systems. This interdisciplinary approach suggests that quantum-theoretic principles may underpin a novel understanding of explainable AI.

Plantation management to restore early and late-successional bird habitat under various climatic and seasonal conditions.

Although agriculture and plantation forestry have decreased natural open habitats and old-growth forests, conservation in managed lands is considered essential for achieving "nature-positive" goals that reverse biodiversity trends from negative to positive. From subboreal to temperate regions, mature conifer plantations with broadleaved trees (BLTs) offer suitable habitats for species preferring mature natural BLT forests, whereas young plantations harbor species depending on early successional (ES) habitats. However, the functional forms of stand age and BLT, and their context dependency, remain unknown. We quantified the effects of stand age (3-98 years), BLT proportion (0%-97.5%), and their interaction on ES and BLT bird species in plantations, as well as the dependency of these effects on regional/seasonal climates. For both groups, we also explored whether plantations could be comparable habitats to BLT-dominated natural forests (stand age: 6-134 years) across Hokkaido (78,420 km2), northern Japan. ES bird species' richness and abundance decreased exponentially with stand age in plantations. This pattern was not evident in natural forests although only two ES stands were surveyed due to the rarity of natural forest harvesting. ES plantations in cooler regions showed higher habitat values, reflecting a climate-dependent species composition. No ES species occurred in winter. Both stand age and BLT proportion increased BLT bird species richness and abundance in a concave manner, except for age in stands with few BLTs. The positive effects of BLT were more evident in younger stands with fewer BLTs. Mature plantations with 25% BLTs supported 62% of breeding BLT bird abundance in old natural forests. In winter, lower regional temperatures weakened the positive effects of stand age and strengthened the positive effects of BLT proportion, reflecting temperature-dependent habitat selection across species. Our results suggest that regular plantation harvesting can play a critical role in restoring ES bird species across Hokkaido. To conserve BLT bird species, retaining even small amounts of BLTs within plantations may be more cost-effective than long-rotation plantation forestry or only protecting existing natural old-growth forests. Our study shows that on-site conservation within plantations across regions and seasons, when coupled with nature reserve management, can contribute to restoring biodiversity.

A conditional random field based approach for high-accuracy part-of-speech tagging using language-independent features

Part-of-speech (POS) tagging is the process of assigning tags or labels to each word of a text based on the grammatical category. It provides the ability to understand the grammatical structure of a text and plays an important role in many natural language processing tasks like syntax understanding, semantic analysis, text processing, information retrieval, machine translation, and named entity recognition. The POS tagging involves sequential nature, context dependency, and labeling of each word. Therefore it is a sequence labeling task. The challenges faced in Urdu text processing including resource scarcity, morphological richness, free word order, absence of capitalization, agglutinative nature, spelling variations, and multipurpose usage of words raise the demand for the development of machine learning automatic POS tagging systems for Urdu. Therefore, a conditional random field (CRF) based supervised POS classifier has been developed for 33 different Urdu POS categories using the language-independent features of Urdu text for the Urdu news dataset MM-POST containing 119,276 tokens of seven different domains including Entertainment, Finance, General, Health, Politics, Science and Sports. An analysis of the proposed approach is presented, proving it superior to other Urdu POS tagging research for using a simpler strategy by employing fewer word-level features as context windows together with the word length. The effective utilization of these features for the POS tagging of Urdu text resulted in the state-of-the-art performance of the CRF model, achieving an overall classification accuracy of 96.1%.

Plant-microbe interactions in tropical and subtropical ecosystems.

Microbes regulate many dimensions of plant performance with multiscale implications for plant fitness, competition, coexistence, and ecosystem functioning. Yet, this fascinating and diverse arena of study has been limited to a few thematic areas, ecosystems, and regions. In particular, despite growing evidence that microbes may be critical players in the dynamics of plant communities in tropical and subtropical ecosystems, these regions remain poorly represented in studies of plant-microbe interactions. Such geographical gaps limit our ability to draw general inferences to comprehend how microbial effects on plant community dynamics may vary with context and, by extension, respond to global environmental change. In this special section of the American Journal of Botany, we bring together a diverse set of research on plant-microbe interactions from tropical and subtropical ecosystems. These papers explore intraspecific variation of soil microbial communities, the context dependency of host-specific assembly of microbial communities on plants, and the new and exciting frontier of the microbiome of epiphytic plants. We hope that this compilation will fuel deeper forays into the many dimensions of plant-microbe interactions in megadiverse tropical and subtropical forests.

DDI-GPT: Explainable Prediction of Drug-Drug Interactions using Large Language Models enhanced with Knowledge Graphs.

Identifying potential drug-drug interactions (DDIs) before clinical use is essential for patient safety yet remains a significant challenge in drug development. We presented DDI-GPT, a deep learning framework that predicts DDIs by combining knowledge graphs (KGs) and pre-trained large language models (LLMs), enabling early detection of potential drug interactions. We demonstrated that DDI-GPT outperforms current state-of-the-art methods by capturing contextual dependencies between biomedical entities to infer potential DDIs. Through feature attribution methods, we show that our explainable deep learning (DL) models enhance the quality of explanations on the pathways and interactome networks. Using TwoSIDES as our benchmark dataset, DDI-GPT achieved the best performance of 0.964 in AUROC compared with existing DL methods. We also applied DDI-GPT to perform zero-shot prediction on 9,480 DDI records, encompassing 442 distinct drugs from the FDA Adverse Event Reporting System. DDI-GPT can attain a high accuracy of in 0.84 AUROC, with an improvement of 14% compared to the best previously published method. We explored model interpretations on predicted DDIs involving Bruton's tyrosine kinase (BTK) inhibitors and uncovered CYP3A-enriched signals underlying the contaminant use of BTK inhibitors with other drugs leading to toxicity. Altogether, DDI-GPT, implemented as both a web server platform and a software package, identifies DDI events and offers a deep learning tool for drug safety use with expandable features.