Ferris State University

The digitization of healthcare has been accompanied by a rapid expansion of electronic health records (EHRs); however, a significant proportion of critical patient data, specifically medication regimens, remains entrapped within unstructured clinical narratives. The inability to seamlessly compute this data hinders advancements in pharmacovigilance, clinical decision support, and population health management. This study presents a comprehensive, rigorous evaluation of the feasibility of deploying Large Language Models (LLMs) to automate the extraction of structured dosage information (Dose, Daily Frequency, Duration) from outpatient antimicrobial clinical notes sourced from the Collaboration to Harmonize Antimicrobial Registry Measures (CHARM) registry. We scrutinized the performance of five distinct open-weight architectures, namely GPT-OSS:20B, Gemma 2:9B, Mistral 7B, Qwen3:14B and Llama 3.2, across both Zero-Shot and Retrieval Augmented Generation (RAG)-based Few-Shot prompting paradigms. Our analysis reveals a fundamental architectural trade-off: the reasoning-optimized GPT-OSS:20B dominates the zero-shot landscape (F1 > 0.90) by leveraging abstract schema understanding, whereas the instruction-tuned Gemma 2:9B excels in the few-shot setting (F1 ~ 0.99), effectively utilizing examples as guardrails to surpass larger models. Conversely, smaller models (Mistral, Llama) exhibit a prohibitive “hallucination barrier,” rendering them unsafe for unsupervised clinical application. Furthermore, we identify “Inconsistent Unit Handling” and “Complex Temporal Logic” as persistent failure modes that resist simple scaling laws. This report provides a definitive framework for selecting model architectures based on the availability of few-shot examples and highlights the necessity of dynamic RAG strategies to achieve production-grade reliability in medical informatics.

Doxorubicin, a widely used chemotherapy drug, is produced by Streptomyces peucetius ATCC27952. The biosynthesis relies on the cytochrome P450 monooxygenase DoxA, which catalyzes three consecutive late-stage oxidation steps. However, conversion from daunorubicin to doxorubicin is inefficient, necessitating semi-synthetic industrial manufacturing. Here, we address key limitations in DoxA catalysis. We identify the natural redox partners ferredoxin Fdx4 and ferredoxin reductase FdR3 by transcriptomic analysis. We discovered the vicinal oxygen chelate family protein DnrV to prevent product inhibition by binding doxorubicin. Structural analysis of DoxA and density functional theory (DFT) calculations reveal that inefficient C14 hydroxylation results from the unfavorable anti-conformation of the methyl ketone side chain of daunorubicin. We harness these advances for rational strain engineering, leading to an 180% increase in doxorubicin yields and an improved production profile. This study provides singular insights into enzymatic constraints in anthracycline biosynthesis and facilitates cost-effective manufacturing to meet the growing global demand for doxorubicin.

The Multistate Pharmacy Jurisprudence Examination (MPJE), despite its name, does not facilitate multistate pharmacy practice, instead requiring pharmacists to pass separate, state-specific law examinations-a process that is increasingly seen as outdated and inefficient. The proposed Uniform MPJE, targeted to launch in 2026, aims to standardize pharmacy law testing nationwide. This article examines the rationale behind the MPJE, questioning the necessity of any pharmacy law examination in an era of technological advancement, evolving regulatory models, and interprofessional parity. With some states already eliminating the MPJE without evidence of public harm, eliminating any law examination may be more aligned with the future of pharmacy practice rather than adopting a new examination.

Social networking sites, like Twitter, host huge amounts of textual data. That content can be analyzed by means of text mining, machine learning, and natural language processing – each of which requires a suitable text representation. In many applications, a Term-Document Matrix (TDM) representation is used, which ignores term order and dependencies. To address this simplification, we propose a hybrid solution, GB-CW-TDM, that combines a new Graph-Based tweet representation model including Centrality and Weight features, along with TDM. To evaluate our model, we use two datasets related to heart attacks. We classify the tweets into either informational or non-informational classes in relation to the context of heart attack. Results show that GB-CW-TDM improves accuracy by up to 27% relative to baseline and state-of-the-art approaches.