Zhisheng Huang

Multimorbidity risk identification in elderly health consultation scenarios is challenging because chronic disease history and medication exposure may interact through complex causal pathways. Existing LLM-, RAG-, and graph-based retrieval methods often rely on semantic relevance or graph connectivity, which may be insufficient for identifying user-specific risk mechanisms. This study proposes RACPR, a generative AI-enhanced risk-aware causal path re-ranking framework for interpretable multimorbidity risk identification. RACPR ranks candidate causal paths by integrating user-entity alignment, causal coherence, risk contribution, and path length control and uses generative AI to transform selected paths into readable, path-grounded explanations. To support controlled algorithmic evaluation, we constructed a normalized benchmark of 1002 elderly multimorbidity consultation cases covering diabetes, hypertension, and chronic kidney disease. The benchmark and supporting knowledge graph were derived from publicly available biomedical and health information resources, including PubMed abstracts, guideline and review sources, DrugBank medication-safety evidence, and MedlinePlus-based terminology. Under a leakage-controlled setting, only age, diagnosed diseases, and medication exposures were used as model-accessible inputs, while abnormal indicators, support paths, and rationales were reserved for evaluation. On the test set, RACPR achieved an Accuracy of 0.659, a Precision of 0.602, a Recall of 0.938, an F1-score of 0.733, and an AUC of 0.777. Ablation analysis showed that removing the risk-aware component reduced AUC from 0.777 to 0.428. Benchmark-level explanation evaluation further showed improvements in path consistency, path hit rate, and health-oriented plausibility. These findings indicate that risk-aware causal path re-ranking improved risk ranking and explanation grounding relative to the evaluated baselines under the controlled benchmark setting.

Bone regeneration is a complex, tightly regulated process that is central to orthopedics, dentistry, and regenerative medicine. Research has traditionally emphasized osteogenesis, osteoinduction, and osteoconduction; more recently, angiogenic–osteogenic coupling and immune regulation have emerged as equally critical. Rapid growth in the number of relevant cell types, in vivo models, and biomaterial strategies—together with an expanding literature—makes it increasingly difficult to synthesize evidence across studies. Here, we present the Knowledge Graph of Bone Regeneration (KGBR), an AI-friendly framework with a literature-processing workflow that converts literature-derived information into a knowledge-graph data model. Using ontology-aligned semantic search, KGBR supports the exploration of bone–immune interactions and provides a structured synthesis of mechanisms and dominant themes in bone regeneration research. By representing the literature as a single interconnected semantic network, KGBR facilitates integration of existing knowledge and supports hypothesis generation. We anticipate that the KGBR data model will provide a foundation for knowledge-based analytics and decision support in bone regeneration.

Astrocytes play a crucial role in brain health and are implicated in various aspects of Alzheimer’s disease (AD) progression. However, the mechanism underlying astrocyte autophagy and pyroptosis in AD remains unclear. To advance this knowledge, we constructed a knowledge graph of AD and dementia (AdDKG) by integrating literature-derived data with the Systematized Nomenclature of Medicine – Clinical Terms (SNOMED CT). The core knowledge base, addkg , was created through named entity recognition and semantic annotation of 172 283 articles using a concept identifier (CI-er) tool. AdDKG knowledge graph extends addkg knowledge base by incorporating SNOMED CT to standardize entities and enrich their relations and attributes. To validate the usefulness of AdDKG, we designed a case study to investigate the roles of astrocyte autophagy and pyroptosis in AD. The query results demonstrated that AdDKG precisely retrieves and prioritizes highly relevant articles, enabling focused investigation of mechanistic relations that would otherwise require extensive manual curation. Further detailed analysis revealed that abnormal autophagy, pyroptosis-induced inflammation, and their interplay are pivotal in AD progression. These findings highlight the utility and efficiency of AdDKG as a research tool for AD research, through which we elucidated the autophagy-pyroptosis interaction in astrocytes as a critical AD mechanism.

Non-suicidal self-injury (NSSI) refers to intentionally harming one’s own body tissue without intending to die or cause injury to others. Common forms of NSSI include cutting, burning, scratching, hitting oneself, and pulling out hair. There are various etiology for NSSI behaviour, which ranges from that of emotional dysregulation to those of cognitive factors. We are particularly interested in the development of a Cognition–Emotion–Behaviour model of NSSI. However, there exist various states concerning cognition (713 functions), emotion (558 states), and NSSI behaviors (81 ones), according to the well-known medical terminologies SNOMED CT. It is non-trivial to know which cognition functions and which emotion states would lead to which kinds of NSSI behaviours. In this paper, we propose a Cognition–Emotion–Behaviour model of NSSI by using an approach of Knowledge Graphs with semantic analysis on the literature mining of NSSI. We will show how cognition, emotion, and NSSI behaviours can be explored by using this knowledge graph approach.