搜索结果：Hugging

Multimodal approaches are increasingly leveraged for integrating omics data with textual biological knowledge. Yet there is still no accessible, standardized framework that enables systematic comparison of omics representations with different text encoders within a unified workflow. We present mmContext, a lightweight and extensible multimodal embedding framework built on top of the open-source Sentence Transformers library. The software allows researchers to train or apply models that jointly embed omics and text data using any numeric representation stored in an AnnData.obsm layer and any text encoder available in Hugging Face. mmContext supports integration of diverse biological text sources and provides pipelines for training, evaluation, and data preparation. We train and evaluate models for a RNA-Seq and text integration task, and demonstrate their utility through zero-shot classification of cell types and diseases across four independent datasets. By releasing all models, datasets, and tutorials openly, mmContext enables reproducible and accessible multimodal learning for omics-text integration. Pretrained checkpoints and full source code for our custom MMContextEncoder are available on Hugging Face huggingface.co/jo-mengr. The Python package github.com/mengerj/mmcontext provides the model implementation and training and evaluation scripts for custom training. The releases for the publication can be accessed via zenodo: adata_hf_datasets: doi.org/10.5281/zenodo.19185217 and mmContext: doi.org/10.5281/zenodo.19185493.

Computational prediction of blood-brain barrier (BBB) permeability has emerged as a vital alternative to traditional experimental assays, which are often resource-intensive and lowthroughput to meet the demands of early-stage drug discovery. While early machine learning approaches have shown promise, integration of traditional chemical descriptors with deep learning embeddings remains an underexplored frontier. In this paper, we introduce TITAN-BBB , a multi-modal deep-learning architecture that utilizes tabular, image, and text-based features and combines them using attention mechanisms. To evaluate, we aggregated multiple literature sources to create the largest BBB permeability dataset to date, enabling robust training for both classification and regression tasks. Our results demonstrate that TITAN-BBB achieves 86.5% of balanced accuracy on classification tasks and 0.436 of mean absolute error for regression, outperforming the state-of-the-art by 3.1 percentage points in balanced accuracy and reducing the regression error by 20%. Our approach also outperforms state-of-the-art models in both classification and regression performance, demonstrating the benefits of combining deep and domain-specific representations. The source code is publicly available at https://github.com/pcdslab/TITAN-BBB . The inference-ready model is hosted on Hugging Face at https://huggingface.co/SaeedLab/TITAN-BBB , and the aggregated BBB permeability datasets are available at https://huggingface.co/datasets/SaeedLab/BBBP .

Foundational models that learn the "language" of molecules are essential for accelerating the material and drug discovery. These self-learning models can be trained on a large number of unlabelled molecules, enabling applications like property prediction, molecule design (de novo generation, optimization), and screening for specific functions. However, existing molecular language models are built upon first-generation transformer architectures and are pretrained using masked language modeling, a generic token-level objective that is agnostic to physicochemical and structural molecular properties. Here we introduce MolDe-BERTa , a structure-informed self-supervised molecular encoder that leverages a byte-level Byte-Pair Encoding (BPE) tokenization strategy. MolDeBERTa is pretrained on up to 123 million SMILES molecules from PubChem, representing one of the largest publicly available SMILES-based corpora. To achieve this, we introduce three novel pretraining objectives designed to inject strong inductive biases for molecular properties and structural similarity directly into the latent space, resulting in reduced gap between linguistic chemical representations and physical molecular properties. The model was then systematically investigated across three architectural scales, two dataset sizes, and five distinct pretraining objectives. MolDeBERTa when evaluated on 9 downstream MoleculeNet benchmarks outperformed existing masked language models, achieving up to a 16% reduction in regression error and improvements of up to 3.0 ROC-AUC points on classification benchmarks. MolDeBERTa advances unsupervised encoder-based foundational models at scale both for pretraining data and downstream evaluation, enabling data-efficient chemistry-informed representation learning. The source code is publicly available at https://github.com/pcdslab/MolDeBERTa , and Hugging Face at https://huggingface.co/collections/SaeedLab/moldeberta . All the pretraining datasets are available at https://huggingface.co/datasets/SaeedLab/MolDeBERTa .

Face anti-spoofing (FAS) has become a crucial component in securing face recognition systems against presentation attacks, such as printed photos, replay videos, and 3D masks. While recent advances have improved generalization to unseen spoofing attempts, many existing methods remain black-box models that provide binary decisions without interpretable reasoning. In this paper, we investigate explainable face anti-spoofing from a supervision-centric perspective, using a vision-language model (VLM) to analyze how natural language explanations influence model behavior. To enable this study under controlled conditions, we construct an explanation-augmented benchmark by enriching four standard FAS datasets-MSU-MFSD, CASIA-FASD, Replay-Attack, and OULU-NPU-with both vanilla and reasoning-structured captions generated via the GPT-4o API. We further adopt a dual-objective training strategy that combines spoof classification loss with explanation generation loss, allowing us to examine the effect of explanation-based supervision while keeping the backbone architecture fixed. Through extensive cross-dataset evaluations, we show that reasoning-style captions can enhance detection performance and domain generalization in many settings, while also introducing inductive biases that may degrade performance when emphasized cues are misaligned with unseen attack types. These findings suggest that explanations in FAS should be viewed not only as interpretable outputs, but also as controllable training signals that shape generalization behavior. To support reproducibility, we publicly release the explanation annotations and associated metadata-excluding all face images-via a Hugging Face repository at https://huggingface.co/datasets/DescriptiveFAS/MCIO_public.

Reliable predictions of protein-protein binding affinities are essential for molecular biology and therapeutic discovery. However, most computational methods rely on three-dimensional structural models, which are often unavailable for many complexes. We introduce BindPred, a structure-agnostic input framework that predicts affinities directly from amino acid sequences by combining embeddings from large protein language models with gradient boosting trees. On the protein-protein binding (PPB)-Affinity benchmark, which comprises 11 919 diverse complexes, BindPred achieves a Pearson correlation coefficient of 0.86 in random split five-fold cross-validation. Ablation analysis indicates that evolutionary embeddings alone capture most of the predictive signals, while augmenting with physics-based energy terms from PyRosetta and BindCraft increases the correlation only by 0.01. A more stringent protein-level split that places entire protein families (wild-type and all mutants) exclusively in either training or testing sets, resulting in only a modest decline in performance, demonstrating robust generalization to novel interaction pairs. Because BindPred operates exclusively on sequence input, it enables rapid inference [approximately 3 million complexes per GPU (T4) hour], making proteome-scale screening computationally feasible. The pretrained model and inference pipeline are available in a Google Colab notebook: BindPred Colab notebook. The training dataset, code, and model weights are available on the hugging face: https://huggingface.co/hbp5181/BindPred.

Standardizing medication concepts across heterogeneous vocabularies is essential for interoperable analytics and observational research. In the Observational Medical Outcomes Partnership (OMOP) Common Data Model, local drug codes must be mapped to standardized RxNorm concepts, but automated mapping is challenging because drug strings encode clinically critical attributes, including strength, dosage form/route, release characteristics, and brand. We propose THIRAWAT (Terminology Harmonization using Late-Interaction Reranker With Alignment-tuned Transformers), a fine-tuned ColBERTv1 late-interaction reranker, and embed it within THIRAWAT Mapper, a retrieval-reranking pipeline with deterministic tie-breaking and stable ordering. Candidate generation used approximate nearest-neighbor retrieval with a bi-encoder (SapBERT-XLMR or BioLORD-2023). Candidates were reranked by THIRAWAT models that were fine-tuned using one-sided MaxSim and scored at inference using our adapted Bidirectional MaxSim (BiMaxSim) pooling. Finally, a deterministic tie-breaker extracted clinically salient cues, including strength, dosage form/route, release characteristics, and bracketed brand annotations, to resolve near-ties reproducibly. We evaluated three mapping settings: Branded Drugs, Clinical Drugs, and Thai Medicines Terminology (TMT). Using SapBERT-XLMR retrieval with THIRAWAT-Sap-BERT reranking and deterministic tie-breaking, THIRAWAT Mapper achieved MRR@100 values of 0.954 (95% confidence interval [CI], 0.921-0.983), 0.898 (95% CI, 0.866-0.925), and 0.912 (95% CI, 0.891-0.931), outperforming a lexical term frequency-inverse document frequency baseline (0.491, 0.216, and 0.143, respectively). Hits@1 improved to 0.942 (95% CI, 0.899-0.978), 0.859 (95% CI, 0.817-0.898), and 0.868 (95% CI, 0.838-0.896), respectively. BiMaxSim and deterministic tie-breaking improved drug mapping to RxNorm while preserving an efficient runtime profile and stable ordering. Overall, THIRAWAT Mapper offers a pragmatic combination of learned semantic matching and deterministic lexical constraints. Models and code are available on Hugging Face (https://huggingface.co/collections/sidataplus/thirawat) and GitHub (https://github.com/sidataplus/THIRAWAT-mapper).

Despite the widespread success of combination antiretroviral therapy in reducing overall HIV morbidity and mortality, people who inject drugs (PWID) continue to experience elevated risk of infection and suboptimal care outcomes. Addressing knowledge gaps related to HIV acquisition and transmission is essential for reducing incident cases of HIV. We assessed HIV transmission misconceptions (e.g., HIV can be transmitted through sharing cigarettes, hugging an infected person) among HIV-negative PWIDs recruited through two syringe service programs in Massachusetts between 2020 and 2025. Participants completed 10 items adapted from the International AIDS Questionnaire, with responses categorized into three groups: 'many misconceptions' (0-7 questions answered correctly), 'few misconceptions' (8-9 correct), and 'no misconceptions' (10 correct). Of 185 participants, 79% held at least one misconception. The most common misconception was that HIV could be transmitted through mosquitoes (55% incorrect). The least prevalent misconceptions related to condom use, drug injection "works" (e.g., cookers, cottons), and hugging (≥ 90% answered correctly). We also observed misconceptions related to syringe sharing (13% incorrect) and mother-to-child transmission (17% incorrect; no difference between women and men). Sociodemographic characteristics did not differ significantly across misconception categories. Participants who reported a recent overdose represented 27% of the full sample but 45% of the 'no misconceptions' subgroup. Misconceptions present since the early era of the HIV epidemic remain prevalent among PWID in this region, despite access to robust harm reduction and HIV prevention services. Findings highlight the importance of dispelling long-standing myths, strengthening education about preventive perinatal transmission, and reinforcing knowledge of injection-specific transmission.

Glycans are highly diverse biological sequences, but their functional understanding has lagged behind proteins and nucleic acids. Many glycans remain ambiguously annotated, limiting computational analyses. Existing computational approaches are primarily graph-based, capturing local structural features but struggling to model global patterns and incomplete sequences. We present GlycanGT, a graph-transformer-based pretrained model for glycans. Glycans were represented as graphs of monosaccharides and glycosidic bonds, and the model was pretrained using a masked language modeling objective. GlycanGT demonstrated higher performance than existing methods across 8 benchmark tasks (e.g., 0.844 AUPRC for immunogenicity classification), and its embeddings formed biologically relevant clusters that recovered known N- and O-glycan categories. Moreover, GlycanGT accurately proposed candidates for ambiguous sequences, maintaining >80% top-5 accuracy for both monosaccharide and glycosidic bond predictions under high masking levels. The source code used in this study is available at https://github.com/matsui-lab/GlycanGT and archived on Zenodo (DOI: 10.5281/zenodo.18636040); pretrained model weights are provided via Hugging Face (https://huggingface.co/Akikitani295/GlycanGT). matsui.yusuke.d4@f.mail.nagoya-u.ac.jp. Supplementary data are available at Bioinformatics online.

The emergence of novel viral pathogens poses critical threats to global health, yet current computational approaches for viral risk assessment are predominantly virus-specific and require extensive retraining for each new threat. Computational methods for rapid characterization of emerging viruses across multiple epidemiologically relevant dimensions-pathogenicity, host tropism, and transmissibility-are urgently needed to inform public health responses and guide experimental prioritization. We present HViLM (Human Virome Language Model) , the first foundation model for panviral genomic analysis through continued pre-training of DNABERT-2 on 5 million non-redundant viral sequences (MMseqs2-clustered from 25 million chunks at 80% identity) spanning 9,000 species across 45+ viral families from the VIRION database. We introduce the Human Virome Understanding Evaluation (HVUE) benchmark comprising seven curated datasets across three prediction tasks: pathogenicity classification, host tropism prediction, and transmissibility assessment. Through parameter-efficient fine-tuning with LoRA, HViLM achieves state-of-the-art performance with average accuracies of 95.32% for pathogenicity, 96.25% for host tropism, and 97.36% for transmissibility assessment. The model demonstrates robust cross-family generalization, substantially outperforming sequence-similarity baselines and general genomic foundation models. Attention-based interpretability analysis reveals that HViLM captures biologically meaningful pathogenicity determinants through molecular mimicry of host regulatory elements, including convergent evolution of eight independent sequences targeting Interferon Regulatory Factor 1 (Irf1) for immune evasion. The HVUE benchmark datasets, training scripts, and complete implementation are publicly available at https://github.com/duttaprat/HViLM . Pre-trained HViLM-base model weights and fine-tuned task-specific variants are available on Hugging Face at https://huggingface.co/duttaprat/HViLM-base . Supplementary data are available online.

Bioactive peptides exhibit immense potential in pharmaceutical and food science domains, with antioxidant peptides (AOPs) garnering significant attention for their roles in scavenging free radicals. However, traditional discovery methods are inefficient and costly. This study introduces RLAnOxPeptide, an integrated computational framework that merges machine learning and reinforcement learning for the efficient prediction and de novo design of AOPs. The framework initially establishes a high-precision predictor, RLP-T5Pred, based on the ProtT5 model via a 'protein-to-peptide' knowledge transfer strategy. By employing label smoothing and logit penalty regularization, it achieves state-of-the-art accuracy (AUC-ROC: 0.9692) and robust calibration. The second component is the generator, RLP-T5Gen, which is trained in an iterative 'Yin-Yang' loop combining supervised learning (to maintain sequence syntax) and reinforcement learning (to drive innovation). Guided by RLP-T5Pred serving as a fixed evaluator and a multi-objective reward function, the generator efficiently designs novel AOPs with high predicted activity. We experimentally validated the framework by synthesizing 17 designed peptides. Most candidates demonstrated potent radical scavenging abilities in chemical assays (DPPH and ABTS), leading to the selection of the top five candidates for cellular validation. In a t-BHP-induced HepG2 cell model, peptides Pep4, Pep5, Pep10, and Pep11 exhibited significant protective effects against oxidative damage. Consequently, the RLAnOxPeptide framework provides a powerful, experimentally verified paradigm for accelerating the discovery of novel antioxidant peptides. The datasets generated and/or analysed during the current study, along with model outputs and representative peptide sequences, have been deposited in a public repository. The RLAnOxPeptide framework source code is available at GitHub: https://github.com/changshh/RLAnOxPeptide. An archival snapshot of the code used to perform the experiments described in this manuscript has been deposited in Zenodo with the DOI: 10.5281/zenodo.20078425. An interactive online demonstration is also available via Hugging Face Spaces: https://huggingface.co/spaces/chshan/RLAnOxPeptide.

This article introduces the DeciViT-Knee 2025 protocol, a hybrid deep-learning workflow that combines Vision Transformers (ViT-B/16), Bayesian Decision Trees, and a third Fuzzy Inference Layer for the early and interpretable detection of Knee Osteoarthritis (KOA). The process starts by setting up a GPU-enabled environment in PyTorch 2.x and Hugging Face Transformers. Then, it gets radiographic and MRI datasets from the Osteoarthritis Initiative (OAI) and the Multicenter Osteoarthritis Study (MOST). Before the fine-tuned ViT-B/16 model processes the data to obtain 768-dimensional image embeddings, it is curated through contrast enhancement, artefact removal, and normalization. We use Bayesian Decision Trees to analyze these embeddings and provide us with calibrated probabilistic classifications. The fuzzy-inference layer incorporates linguistic reasoning to express diagnostic uncertainty in terms such as low, medium, or high risk. The model's performance surpasses that of the CNN, ResNet-50, and ViT-only baselines, achieving an accuracy of 92.4%, an AUC of 0.964, an F1 score of 0.891, and a Brier score of 0.088. The protocol demonstrates how combining transformer-based global feature extraction, Bayesian probabilistic reasoning, and fuzzy interpretability can create a clearer, reproducible, and clinician-friendly framework for diagnosing musculoskeletal imaging.

Posterior fibre-reinforced bridges (FRBs) offer a conservative and cost-effective alternative for tooth replacement, particularly in cases where implants or conventional prostheses are not feasible. Despite their clinical potential, inconsistent techniques and the lack of standardised protocols have limited broader adoption. The aim of this article is to present a reproducible, clinically grounded protocol for direct posterior FRB fabrication, integrating strategic cavity design, appropriate fibre selection, and biomechanically guided fibre placement. The technique incorporates the Hugging Mylar Band Technique, which allows single step pontic fabrication and enhances cervical surface smoothness in areas that are difficult to polish, with potential implications for hygiene maintenance and workflow efficiency. The protocol is grounded in current scientific literature and refined through extensive clinical experience over more than a decade, offering a structured, aesthetic, and biologically conservative restorative solution. Although long-term survival outcomes for posterior FRBs remain variable, this protocol aims to improve clinical consistency by standardising the key biomechanical principles involved.

Korean American parents express warmth indirectly (e.g., through devoted attention), which can be contrasted with explicit expressions of warmth (e.g., stating love, hugging). This study investigated three distinct forms of indirect warmth. Thoughtfulness involves parents anticipating and meeting children's needs before they are expressed. Devotion to education concerns prioritizing a supportive learning environment over parental interests. Guan behavior involves communicating clear standards to children, in a manner sensitive to their abilities. The present longitudinal study examined, in a path analysis, whether Time 1 (T1) Korean American adolescents' perceptions of the three separate forms of indirect warmth were positively predicted by T1 maternal reports of supportiveness and demandingness or guan ideology. Guan ideology predicted devotion, β = 0.08, p = 0.048, as well as guan behavior, β = 0.11, p = 0.008. Supportiveness predicted thoughtfulness, β = 0.11, p = 0.005. The study also examined whether the T1 variables predicted T2 adolescent outcomes. T1 thoughtfulness predicted T2 prosocial behavior, β = 0.15, p = 0.007, and relationship quality, β = 0.05, p = 0.001. T2 GPA was predicted by T1 supportiveness, β = 0.12, p = 0.048, and devotion, β = 0.18, p = 0.014. There was evidence for full mediation from supportiveness via thoughtfulness for relationship quality, β = 0.02, p = 0.029, and prosocial behavior β = 0.02, p = 0.052. Because supportiveness had its influence on the latter two variables via thoughtfulness, those who work with Korean American families may want to consider addressing thoughtfulness directly.

This paper presents a novel synthetic, automatically annotated dialogue-style dataset derived from narrative short stories, formatted as Speaker(Emotion): Dialogue. The corpus comprises 2,588 stories containing 27,074 dialogue utterances attributed to 2,582 unique characters, each annotated with one of six emotion categories (Joy, Sadness, Anger, Fear, Disgust, Surprise). Built from a Hugging Face story corpus, the dataset captures speaker attribution and emotional context across dialogues. To accurately identify speakers and their genders, we employed FastCoref with masked dialogue processing, followed by BookNLP for enhanced entity resolution. Emotional labels were assigned using a Bi-GRU-based emotion classifier, while representative story-level keywords were extracted using Mistral-7B via prompt engineering. The resulting dataset pairs keywords with fully-annotated emotional dialogues, enabling keyword-conditioned narrative generation. This dataset serves as a foundation for fine-tuning generative language models and was further used to generate narrations with emotional TTS using ElevenLabs, showcasing the practical utility of our annotations for emotionally expressive storytelling.

A hug is a simple act of human connection, yet for many individuals with abdominal obesity, it can be physically challenging and emotionally uncomfortable. This paper examines how abdominal obesity affects both health and social interactions, using hugging as an illustrative example. In collectivist cultures such as Bangladesh, physical touch reinforces trust and family bonds. Obesity-related limitations can reduce closeness; promote avoidance and lower self-confidence, often compounded by social stigma. Abdominal obesity is also associated with increased risks of cardiovascular disease, diabetes, and premature mortality. Its prevalence is driven by sedentary lifestyles, processed food consumption, and environmental factors, further exacerbated during the COVID-19 pandemic. Addressing abdominal obesity requires holistic strategies, including improved nutrition, regular physical activity, stress management, and community support, to restore comfort, rebuild confidence, and strengthen personal relationships and community cohesion.