Motivation Pharmacovigilance relies on accurate extraction of structured biomedical entities and their semantic relationships from scientific literature. However, most biomedical information extraction systems address named entity recognition (NER) and relation extraction as separate tasks trained on corpus-specific architectures, limiting scalability and cross-task knowledge sharing. Recent developments in instruction-tuned Large Language Models (LLMs) offer a promising alternative through unified generative extraction, but robust schema-grounded multitask adaptation for biomedical extraction is still understudied. Methods This study proposes a unified multitask instruction-tuned LLM framework that jointly performs biomedical NER and relation extraction across three benchmark corpora to identify chemical, disease, drug entities, as well as chemical-disease relations, drug-adverse event relations, and drug-drug interactions. Two general LLMs, Llama-3.2-3B-Instruct and Qwen3-8B, were fine-tuned using Low-Rank Adaptation (LoRA) under a shared generation interface that extracts both entity pairs and their underlying relation. Zero-shot and fine-tuned configurations were evaluated across all the tasks on their respective held-out test sets. Results Parameter-efficient fine-tuning substantially improved both entity and relation extraction performance across all tasks and model families. Fine-tuned Qwen3-8B achieved the strongest overall performance with 89.42% micro-averaged entity F1 and 62.32% micro-averaged relation F1. Fine-tuned Llama-3.2-3B achieved 87.63% entity F1 and 58.42% relation F1 despite its substantially smaller parameter count, outperforming the zero-shot 8B model on both tasks. Fine-tuning also reduced structured JSON parse failures from 23.5% to 0.11%, demonstrating stable schema internalization during supervised adaptation. Conclusion Schema-grounded multitask instruction tuning with LoRA provides a robust and computationally feasible framework for unified biomedical information extraction across heterogeneous benchmark corpora. The findings further demonstrate that schema-grounded adaptation is substantially more important than model scale alone for reliable extraction of structured biomedical relations. The gap between NER and relation extraction performance motivates future research on explicit negative-relation supervision and ontology-guided relation extraction.
Tumor-associated macrophage (TAM) infiltration is a critical characteristic of triple-negative breast cancer (TNBC) related to drug resistance and poor prognosis. Integrating macrophages into TNBC spheroids is crucial to improve the accuracy of 3D in vitro models that mimic the complexity of the tumor microenvironment (TME) and assess treatment response. However, this remains challenging since the reciprocal effects of these two cell types on each other are not fully understood. In this study, we used the TNBC cell line, MDA-MB-231, and polarized M1-like or M2-like macrophages derived from THP-1 monocytes to establish 3D co-culture spheroids to examine bidirectional interactions between these cells and responses to chemotherapy. Drug efficacy, epithelial-mesenchymal transition (EMT) in cancer cells, macrophage phenotypes, and RNA sequencing, including pathway enrichment analysis, were performed in 3D spheroids. CIBERSORTx deconvolution of RNA sequencing results facilitated the separation of cell types within mixtures to estimate their corresponding cell fractions. We observed that M2 macrophages increased the viability of MDA-MB-231 cells in 3D spheroids, while both M1 and M2 macrophages increased the chemosensitivity of 3D spheroids to doxorubicin and paclitaxel. Interestingly, instead of maintaining their phenotypes, both M1 and M2 macrophages lost some polarization and formed a mixed M1-M2 phenotype when co-cultured with MDA-MB-231 cells in 3D spheroids, a phenomenon further supported by RNA-seq deconvolution analysis. However, the fraction of M1-like macrophages shifting to M2-like was much lower than the fraction of M2-like macrophages shifting to M1-like in the 3D co-cultures. Compared with 2D cultures, an expected mesenchymal transition, numerous differentially expressed genes (DEGs) and various pathways, including both tumor-promoting and tumor-suppressing genes, were observed in 3D spheroid MDA-MB-231 cells. However, both M1- and M2-like macrophages induced only partial EMT phenotype changes of cancer cells in co-cultures. Furthermore, a coexistence of pro-inflammatory and anti-inflammatory DEGs was observed in both M1 and M2-like co-cultured cancer spheroids. In conclusion, our findings present an effective 3D co-culture system of breast cancer cells and integrated macrophages for studying dynamic cellular phenotype changes and reciprocal interactions in a heterogeneous environment to mimic aspects of the TME and enhance the accuracy of preclinical in vitro treatment response studies.
Chen Cheng, Brett A. McGregor, Junguk Hur et al.· PLoS ONE· 0 citations
We use cookies to run the site and, with your consent, for analytics and to show ads.
See our Cookie Policy.
