Douglas Teodoro

Anthony Yazdani, Dimitrios Proios, Hossein Rouhizadeh et al.

Current approaches for clinical information extraction are inefficient in terms of computational costs and memory consumption, hindering their application to process large-scale electronic health records (EHRs). We propose an efficient end-to-end model, the Joint-NER-RE-Fourier (JNRF), to jointly learn the tasks of named entity recognition and relation extraction for documents of variable length. The architecture uses positional encoding and unitary batch sizes to process variable length documents and uses a weight-shared Fourier network layer for low-complexity token mixing. Finally, we reach the theoretical computational complexity lower bound for relation extraction using a selective pooling strategy and distance-aware attention weights with trainable polynomial distance functions. We evaluated the JNRF architecture using the 2018 N2C2 ADE benchmark to jointly extract medication-related entities and relations in variable-length EHR summaries. JNRF outperforms rolling window BERT with selective pooling by 0.42%, while being twice as fast to train. Compared to state-of-the-art BiLSTM-CRF architectures on the N2C2 ADE benchmark, results show that the proposed approach trains 22 times faster and reduces GPU memory consumption by 1.75 folds, with a reasonable performance tradeoff of 90%, without the use of external tools, hand-crafted rules or post-processing. Given the significant carbon footprint of deep learning models and the current energy crises, these methods could support efficient and cleaner information extraction in EHRs and other types of large-scale document databases.

Anthony Yazdani, Hossein Rouhizadeh, David Vicente Alvarez et al.

This paper outlines the performance evaluation of a system for adverse drug event normalization, developed by the Data Science for Digital Health (DS4DH) group for the Social Media Mining for Health Applications (SMM4H) 2023 shared task 5. Shared task 5 targeted the normalization of adverse drug event mentions in Twitter to standard concepts of the Medical Dictionary for Regulatory Activities terminology. Our system hinges on a two-stage approach: BERT fine-tuning for entity recognition, followed by zero-shot normalization using sentence transformers and reciprocal-rank fusion. The approach yielded a precision of 44.9%, recall of 40.5%, and an F1-score of 42.6%. It outperformed the median performance in shared task 5 by 10% and demonstrated the highest performance among all participants. These results substantiate the effectiveness of our approach and its potential application for adverse drug event normalization in the realm of social media text mining.

Rui Yang, Huitao Li, Weihao Xuan et al.

Despite continuous advances in medical technology, the global distribution of health care resources remains uneven. The development of large language models (LLMs) has transformed the landscape of medicine and holds promise for improving health care quality and expanding access to medical information globally. However, existing LLMs are primarily trained on high-resource languages, limiting their applicability in global medical scenarios. To address this gap, we constructed GlobMed, a large multilingual medical dataset, containing over 500,000 entries spanning 12 languages, including four low-resource languages. Building on this, we established GlobMed-Bench, which systematically assesses 56 state-of-the-art proprietary and open-weight LLMs across multiple multilingual medical tasks, revealing significant performance disparities across languages, particularly for low-resource languages. Additionally, we introduced GlobMed-LLMs, a suite of multilingual medical LLMs trained on GlobMed, with parameters ranging from 1.7B to 8B. GlobMed-LLMs achieved an average performance improvement of over 40% relative to baseline models, with a more than threefold increase in performance on low-resource languages. Together, these resources provide an important foundation for advancing the equitable development and application of LLMs globally, enabling broader language communities to benefit from technological advances.

Rui Yang, Matthew Yu Heng Wong, Huitao Li et al.

The rapid growth of medical knowledge and increasing complexity of clinical practice pose challenges. In this context, large language models (LLMs) have demonstrated value; however, inherent limitations remain. Retrieval-augmented generation (RAG) technologies show potential to enhance their clinical applicability. This study reviewed RAG applications in medicine. We found that research primarily relied on publicly available data, with limited application in private data. For retrieval, approaches commonly relied on English-centric embedding models, while LLMs were mostly generic, with limited use of medical-specific LLMs. For evaluation, automated metrics evaluated generation quality and task performance, whereas human evaluation focused on accuracy, completeness, relevance, and fluency, with insufficient attention to bias and safety. RAG applications were concentrated on question answering, report generation, text summarization, and information extraction. Overall, medical RAG remains at an early stage, requiring advances in clinical validation, cross-linguistic adaptation, and support for low-resource settings to enable trustworthy and responsible global use.

Boya Zhang, Alban Bornet, Rui Yang et al.

How do language models use contextual information to answer health questions? How are their responses impacted by conflicting contexts? We assess the ability of language models to reason over long, conflicting biomedical contexts using HealthContradict, an expert-verified dataset comprising 920 unique instances, each consisting of a health-related question, a factual answer supported by scientific evidence, and two documents presenting contradictory stances. We consider several prompt settings, including correct, incorrect or contradictory context, and measure their impact on model outputs. Compared to existing medical question-answering evaluation benchmarks, HealthContradict provides greater distinctions of language models' contextual reasoning capabilities. Our experiments show that the strength of fine-tuned biomedical language models lies not only in their parametric knowledge from pretraining, but also in their ability to exploit correct context while resisting incorrect context.

Félicien Hêche, Sohrab Ferdowsi, Anthony Yazdani et al.

Objective: The objective of this study is to develop a machine learning (ML)-based framework for early risk stratification of clinical trials (CTs) according to their likelihood of exhibiting a high rate of dosing errors, using information available prior to trial initiation. Materials and Methods: We constructed a dataset from ClinicalTrials.gov comprising 42,112 CTs. Structured, semi-structured trial data, and unstructured protocol-related free-text data were extracted. CTs were assigned binary labels indicating elevated dosing error rate, derived from adverse event reports, MedDRA terminology, and Wilson confidence intervals. We evaluated an XGBoost model trained on structured features, a ClinicalModernBERT model using textual data, and a simple late-fusion model combining both modalities. Post-hoc probability calibration was applied to enable interpretable, trial-level risk stratification. Results: The late-fusion model achieved the highest AUC-ROC (0.862). Beyond discrimination, calibrated outputs enabled robust stratification of CTs into predefined risk categories. The proportion of trials labeled as having an excessively high dosing error rate increased monotonically across higher predicted risk groups and aligned with the corresponding predicted probability ranges. Discussion: These findings indicate that dosing error risk can be anticipated at the trial level using pre-initiation information. Probability calibration was essential for translating model outputs into reliable and interpretable risk categories, while simple multimodal integration yielded performance gains without requiring complex architectures. Conclusion: This study introduces a reproducible and scalable ML framework for early, trial-level risk stratification of CTs at risk of high dosing error rates, supporting proactive, risk-based quality management in clinical research.

Anthony Yazdani, Ihor Stepanov, Douglas Teodoro

Biomedical named entity recognition (NER) presents unique challenges due to specialized vocabularies, the sheer volume of entities, and the continuous emergence of novel entities. Traditional NER models, constrained by fixed taxonomies and human annotations, struggle to generalize beyond predefined entity types. To address these issues, we introduce GLiNER-BioMed, a domain-adapted suite of Generalist and Lightweight Model for NER (GLiNER) models specifically tailored for biomedicine. In contrast to conventional approaches, GLiNER uses natural language labels to infer arbitrary entity types, enabling zero-shot recognition. Our approach first distills the annotation capabilities of large language models (LLMs) into a smaller, more efficient model, enabling the generation of high-coverage synthetic biomedical NER data. We subsequently train two GLiNER architectures, uni- and bi-encoder, at multiple scales to balance computational efficiency and recognition performance. Experiments on several biomedical datasets demonstrate that GLiNER-BioMed outperforms the state-of-the-art in both zero- and few-shot scenarios, achieving 5.96% improvement in F1-score over the strongest baseline (p-value < 0.001). Ablation studies highlight the effectiveness of our synthetic data generation strategy and emphasize the complementary benefits of synthetic biomedical pre-training combined with fine-tuning on general-domain annotations. All datasets, models, and training pipelines are publicly available at https://github.com/ds4dh/GLiNER-biomed.

Weihao Xuan, Rui Yang, Heli Qi et al.

Existing large language model (LLM) evaluation benchmarks primarily focus on English, while current multilingual tasks lack parallel questions that specifically assess cross-linguistic reasoning abilities. This dual limitation makes it challenging to comprehensively assess LLMs' performance in the multilingual setting. To fill this gap, we introduce MMLU-ProX, a comprehensive benchmark covering 29 languages, built on an English benchmark. Each language version consists of 11,829 identical questions, enabling direct cross-linguistic comparisons. Additionally, to meet efficient evaluation needs, we provide a lite version containing 658 questions per language. To ensure the high quality of MMLU-ProX, we employ a rigorous development process that involves multiple powerful LLMs for translation, followed by expert review to ensure accurate expression, consistent terminology, and cultural relevance. Building on this, we systematically evaluate 36 state-of-the-art LLMs, including reasoning-enhanced and multilingual-optimized LLMs. The results reveal significant disparities in the multilingual capabilities of LLMs: While they perform well in high-resource languages, their performance declines markedly in low-resource languages, with gaps of up to 24.3%. Through MMLU-ProX, we aim to advance the development of more inclusive AI systems and promote equitable access to technology across global contexts.

Anthony Yazdani, Alban Bornet, Philipp Khlebnikov et al.

Adverse drug events (ADEs) are a major safety issue in clinical trials. Thus, predicting ADEs is key to developing safer medications and enhancing patient outcomes. To support this effort, we introduce CT-ADE, a dataset for multilabel ADE prediction in monopharmacy treatments. CT-ADE encompasses 2,497 drugs and 168,984 drug-ADE pairs from clinical trial results, annotated using the MedDRA ontology. Unlike existing resources, CT-ADE integrates treatment and target population data, enabling comparative analyses under varying conditions, such as dosage, administration route, and demographics. In addition, CT-ADE systematically collects all ADEs in the study population, including positive and negative cases. To provide a baseline for ADE prediction performance using the CT-ADE dataset, we conducted analyses using large language models (LLMs). The best LLM achieved an F1-score of 56%, with models incorporating treatment and patient information outperforming by 21%-38% those relying solely on the chemical structure. These findings underscore the importance of contextual information in ADE prediction and establish CT-ADE as a robust resource for safety risk assessment in pharmaceutical research and development.

Feiyang Li, Yingjian Chen, Haoran Liu et al.

Large Language Models (LLMs) have shown remarkable progress in medical question answering (QA), yet their effectiveness remains predominantly limited to English due to imbalanced multilingual training data and scarce medical resources for low-resource languages. To address this critical language gap in medical QA, we propose Multilingual Knowledge Graph-based Retrieval Ranking (MKG-Rank), a knowledge graph-enhanced framework that enables English-centric LLMs to perform multilingual medical QA. Through a word-level translation mechanism, our framework efficiently integrates comprehensive English-centric medical knowledge graphs into LLM reasoning at a low cost, mitigating cross-lingual semantic distortion and achieving precise medical QA across language barriers. To enhance efficiency, we introduce caching and multi-angle ranking strategies to optimize the retrieval process, significantly reducing response times and prioritizing relevant medical knowledge. Extensive evaluations on multilingual medical QA benchmarks across Chinese, Japanese, Korean, and Swahili demonstrate that MKG-Rank consistently outperforms zero-shot LLMs, achieving maximum 35.03% increase in accuracy, while maintaining an average retrieval time of only 0.0009 seconds.

Rui Yang, Huitao Li, Matthew Yu Heng Wong et al.

Natural language processing (NLP) has been traditionally applied to medicine, and generative large language models (LLMs) have become prominent recently. However, the differences between them across different medical tasks remain underexplored. We analyzed 19,123 studies, finding that generative LLMs demonstrate advantages in open-ended tasks, while traditional NLP dominates in information extraction and analysis tasks. As these technologies advance, ethical use of them is essential to ensure their potential in medical applications.

Boya Zhang, Nona Naderi, Fernando Jaume-Santero et al.

This paper describes the work of the Data Science for Digital Health (DS4DH) group at the TREC Health Misinformation Track 2021. The TREC Health Misinformation track focused on the development of retrieval methods that provide relevant, correct and credible information for health related searches on the Web. In our methodology, we used a two-step ranking approach that includes i) a standard retrieval phase, based on BM25 model, and ii) a re-ranking phase, with a pipeline of models focused on the usefulness, supportiveness and credibility dimensions of the retrieved documents. To estimate the usefulness, we classified the initial rank list using pre-trained language models based on the transformers architecture fine-tuned on the MS MARCO corpus. To assess the supportiveness, we utilized BERT-based models fine-tuned on scientific and Wikipedia corpora. Finally, to evaluate the credibility of the documents, we employed a random forest model trained on the Microsoft Credibility dataset combined with a list of credible sites. The resulting ranked lists were then combined using the Reciprocal Rank Fusion algorithm to obtain the final list of useful, supporting and credible documents. Our approach achieved competitive results, being top-2 in the compatibility measurement for the automatic runs. Our findings suggest that integrating automatic ranking models created for each information quality dimension with transfer learning can increase the effectiveness of health-related information retrieval.

Sohrab Ferdowsi, Nikolay Borissov, Julien Knafou et al.

We consider the hierarchical representation of documents as graphs and use geometric deep learning to classify them into different categories. While graph neural networks can efficiently handle the variable structure of hierarchical documents using the permutation invariant message passing operations, we show that we can gain extra performance improvements using our proposed selective graph pooling operation that arises from the fact that some parts of the hierarchy are invariable across different documents. We applied our model to classify clinical trial (CT) protocols into completed and terminated categories. We use bag-of-words based, as well as pre-trained transformer-based embeddings to featurize the graph nodes, achieving f1-scores around 0.85 on a publicly available large scale CT registry of around 360K protocols. We further demonstrate how the selective pooling can add insights into the CT termination status prediction. We make the source code and dataset splits accessible.

Jenny Copara, Nona Naderi, Julien Knafou et al.

Chemical patent documents describe a broad range of applications holding key reaction and compound information, such as chemical structure, reaction formulas, and molecular properties. These informational entities should be first identified in text passages to be utilized in downstream tasks. Text mining provides means to extract relevant information from chemical patents through information extraction techniques. As part of the Information Extraction task of the Cheminformatics Elsevier Melbourne University challenge, in this work we study the effectiveness of contextualized language models to extract reaction information in chemical patents. We assess transformer architectures trained on a generic and specialised corpora to propose a new ensemble model. Our best model, based on a majority ensemble approach, achieves an exact F1-score of 92.30% and a relaxed F1-score of 96.24%. The results show that ensemble of contextualized language models can provide an effective method to extract information from chemical patents.

Douglas Teodoro, Luc Mottin, Julien Gobeill et al.

Availability of research datasets is keystone for health and life science study reproducibility and scientific progress. Due to the heterogeneity and complexity of these data, a main challenge to be overcome by research data management systems is to provide users with the best answers for their search queries. In the context of the 2016 bioCADDIE Dataset Retrieval Challenge, we investigate a novel ranking pipeline to improve the search of datasets used in biomedical experiments. Our system comprises a query expansion model based on word embeddings, a similarity measure algorithm that takes into consideration the relevance of the query terms, and a dataset categorisation method that boosts the rank of datasets matching query constraints. The system was evaluated using a corpus with 800k datasets and 21 annotated user queries. Our system provides competitive results when compared to the other challenge participants. In the official run, it achieved the highest infAP among the participants, being +22.3% higher than the median infAP of the participant's best submissions. Overall, it is ranked at top 2 if an aggregated metric using the best official measures per participant is considered. The query expansion method showed positive impact on the system's performance increasing our baseline up to +5.0% and +3.4% for the infAP and infNDCG metrics, respectively. Our similarity measure algorithm seems to be robust, in particular compared to Divergence From Randomness framework, having smaller performance variations under different training conditions. Finally, the result categorization did not have significant impact on the system's performance. We believe that our solution could be used to enhance biomedical dataset management systems. In particular, the use of data driven query expansion methods could be an alternative to the complexity of biomedical terminologies.