Exploring the reversal curse and other deductive logical reasoning in BERT and GPT-based large language models.

The "Reversal Curse" describes the inability of autoregressive decoder large language models (LLMs) to deduce "B is A" from "A is B," assuming that B and A are distinct and can be uniquely identified from each other. This logical failure suggests limitations in using generative pretrained transformer (GPT) models for tasks like constructing knowledge graphs. Our study revealed that a bidirectional LLM, bidirectional encoder representations from transformers (BERT), does not suffer from this issue.

Fine-tuning large language models for rare disease concept normalization.

Objective: We aim to develop a novel method for rare disease concept normalization by fine-tuning Llama 2, an open-source large language model (LLM), using a domain-specific corpus sourced from the Human Phenotype Ontology (HPO).

Methods: We developed an in-house template-based script to generate two corpora for fine-tuning. The first (NAME) contains standardized HPO names, sourced from the HPO vocabularies, along with their corresponding identifiers.

D-TERMINED, a phase 1 trial in newly diagnosed high-grade glioma with temozolomide, radiation, and minocycline followed by adjuvant minocycline/temozolomide.

Background: Standard treatment for newly diagnosed high-grade gliomas remains suboptimal. Preclinical data indicate that mesenchymal transition and radiation resistance in glioblastoma are driven by NF-κB and microglia activation, which can be inhibited by minocycline. We assessed the safety and efficacy of minocycline combined with standard radiation and temozolomide in newly diagnosed high-grade gliomas.

GestaltMML: Enhancing Rare Genetic Disease Diagnosis through Multimodal Machine Learning Combining Facial Images and Clinical Texts.

Individuals with suspected rare genetic disorders often undergo multiple clinical evaluations, imaging studies, laboratory tests and genetic tests, to find a possible answer over a prolonged period of time. Addressing this "diagnostic odyssey" thus has substantial clinical, psychosocial, and economic benefits. Many rare genetic diseases have distinctive facial features, which can be used by artificial intelligence algorithms to facilitate clinical diagnosis, in prioritizing candidate diseases to be further examined by lab tests or genetic assays, or in helping the phenotype-driven reinterpretation of genome/exome sequencing data.

Assessing the Utility of Large Language Models for Phenotype-Driven Gene Prioritization in Rare Genetic Disorder Diagnosis.

Phenotype-driven gene prioritization is a critical process in the diagnosis of rare genetic disorders for identifying and ranking potential disease-causing genes based on observed physical traits or phenotypes. While traditional approaches rely on curated knowledge graphs with phenotype-gene relations, recent advancements in large language models have opened doors to the potential of AI predictions through extensive training on diverse corpora and complex models. This study conducted a comprehensive evaluation of five large language models, including two Generative Pre-trained Transformers series, and three Llama2 series, assessing their performance across three key metrics: task completeness, gene prediction accuracy, and adherence to required output structures.

Enhancing phenotype recognition in clinical notes using large language models: PhenoBCBERT and PhenoGPT.

To enhance phenotype recognition in clinical notes of genetic diseases, we developed two models-PhenoBCBERT and PhenoGPT-for expanding the vocabularies of Human Phenotype Ontology (HPO) terms. While HPO offers a standardized vocabulary for phenotypes, existing tools often fail to capture the full scope of phenotypes due to limitations from traditional heuristic or rule-based approaches. Our models leverage large language models to automate the detection of phenotype terms, including those not in the current HPO.

Fine-tuning Large Language Models for Rare Disease Concept Normalization.

Objective: We aim to develop a novel method for rare disease concept normalization by fine-tuning Llama 2, an open-source large language model (LLM), using a domain-specific corpus sourced from the Human Phenotype Ontology (HPO).

Methods: We developed an in-house template-based script to generate two corpora for fine-tuning. The first (NAME) contains standardized HPO names, sourced from the HPO vocabularies, along with their corresponding identifiers.

Classification of integers based on residue classes via modern deep learning algorithms.

Judging whether an integer can be divided by prime numbers such as 2 or 3 may appear trivial to human beings, but it can be less straightforward for computers. Here, we tested multiple deep learning architectures and feature engineering approaches to classifying integers based on their residues when divided by small prime numbers. We found that the ability of classification critically depends on the feature space.

Use of Disease-Modifying Treatments in Patients With Sickle Cell Disease.

Importance: Despite hydroxyurea being an established treatment for sickle cell disease (SCD), it remains underused. The recent approval of the disease-modifying treatments (DMTs) l-glutamine, crizanlizumab, and voxelotor underscores the need to understand the uptake of DMTs in the current treatment landscape.

Objective: To explore characteristics that may be associated with DMT use and to describe observed patterns of yearly DMT use from 2014 to 2021.

Enhancing Phenotype Recognition in Clinical Notes Using Large Language Models: PhenoBCBERT and PhenoGPT.

To enhance phenotype recognition in clinical notes of genetic diseases, we developed two models - PhenoBCBERT and PhenoGPT - for expanding the vocabularies of Human Phenotype Ontology (HPO) terms. While HPO offers a standardized vocabulary for phenotypes, existing tools often fail to capture the full scope of phenotypes, due to limitations from traditional heuristic or rule-based approaches. Our models leverage large language models (LLMs) to automate the detection of phenotype terms, including those not in the current HPO.

Effects of carbohydrate drinks ingestion on executive function in athletes: a systematic review and meta-analysis.

Background: Carbohydrates are often used as boosters for endurance and high-intensity exercise. However, it is unclear whether carbohydrate drinks intake before or during exercise can affect specific domains of cognitive function, such as Executive Function (EF).

Methods: Following the guidance of PRISMA 2020, we searched six major databases including PubMed, WOS, SPORTDiscus, Cochrane, Embase, and Scopus.

Unequal crossing over between CYP11B2 and CYP11B1 causes 11 β -hydroxylase deficiency in a consanguineous family.

Cytochrome P450 (CYP) family CYP11B2/CYP11B1 chimeric genes have been shown to arise from unequal crossing over of the genes encoding aldosterone synthase (CYP11B2) and 11β-hydroxylase (CYP11B1) during meiosis. The activity deficiency or impaired activity of aldosterone synthase and 11β-hydroxylase resulting from these chimeric genes are important reasons for 11β-hydroxylase deficiency (11β-OHD). Here，two patients with pseudoprecocious puberty and hypokalemia hypertension and three carriers in a consanguineous marriage family were studied.

Treatment of open and comminuted mid-distal tibial fractures by bilateral external fixation combined with limited-internal fixation.

Open and comminuted mid-distal fractures often result from high-energy trauma, and a concomitant poor blood supply often leads to skin necrosis, infection, and bone union. To circumvent such complications, we used limited-reduction and bilateral-external fixators to treat open and comminuted mid-distal tibial fractures with compromised soft tissue. A retrospective series of 34 patients who had open and comminuted mid-distal tibial fractures and treated by bilateral-external fixators with limited-internal fixation were analyzed.

Trajectories of Pelvic Floor Symptoms and Support After Vaginal Delivery in Primiparous Women Between Third Trimester and 1 Year Postpartum.

Objectives: The objectives of this study were to describe trajectories of pelvic floor symptoms and support from the third trimester to 1 year postpartum in primiparous women after vaginal delivery and to explore factors associated with their resolution between 8 weeks postpartum and 1 year postpartum.

Methods: Five hundred ninety-seven nulliparous women 18 years or older who gave birth vaginally at term completed the Epidemiology of Prolapse and Incontinence Questionnaire and the Pelvic Organ Prolapse Quantification examination at the third trimester, 8 weeks postpartum, and 1 year postpartum.

Results: At 1 year postpartum, 41%, 32%, and 23% of participants reported stress urinary incontinence, nocturia, and flatus incontinence, respectively, and 9% demonstrated maximal vaginal descent (MVD) ≥ 0 cm.

Effects of Oxytocin for Induction and Augmentation of Labor on Pelvic Floor Symptoms and Support in the Postpartum Period.

Objective: The aim of the study was to determine whether oxytocin for induction or augmentation of labor impacts the incidence or persistence of pelvic floor symptoms and support 5 to 10 weeks after first vaginal delivery.

Methods: Participants in this prospective cohort study were nulliparous women 18 years or older that delivered vaginally at 37 weeks gestation or more and completed the Epidemiology of Prolapse and Incontinence Questionnaire (EPIQ) and the Pelvic Organ Prolapse Quantification examination in third trimester and 5 to 10 weeks postpartum. We compared the incidence and persistence of symptomatic EPIQ domains and worse vaginal support (maximal vaginal descent ≥0 cm) between women who received oxytocin with those that did not (with or without prostaglandin or mechanical methods in both groups).

Procalcitonin and N-Terminal Pro-B-Type Natriuretic Peptide for Prognosis in Septic Acute Kidney Injury Patients Receiving Renal Replacement Therapy.

Background: Sepsis is a complex clinical syndrome leading to severe sepsis and septic shock. It is very common in the intensive care unit with high mortality. Thus, judging its prognosis is extremely important.

Trajectories of lower back, upper back, and pelvic girdle pain during pregnancy and early postpartum in primiparous women.

Objective: Most studies about truncal pain during and after pregnancy focus on low back pain, few prospectively define change in pain, and even fewer evaluate pain in all three major truncal areas: upper back, lower back, and pelvic girdle. Thus, the objective of this prospective cohort study was to describe, in primiparous women delivered vaginally, prevalence rates and severities of upper back, lower back, and pelvic girdle pain during pregnancy and 6-10 weeks postpartum and to describe the trajectory of pain constellations between time points.

Study Design: Participants completed questionnaires at each time point.

MondoA-Mlx transcriptional activity is limited by mTOR-MondoA interaction.

Mammalian target of rapamycin (mTOR) integrates multiple signals, including nutrient status, growth factor availability, and stress, to regulate cellular and organismal growth. How mTOR regulates transcriptional programs in response to these diverse stimuli is poorly understood. MondoA and its obligate transcription partner Mlx are basic helix-loop-helix leucine zipper (bHLHZip) transcription factors that sense and execute a glucose-responsive transcriptional program.

Microfibers fabricated by non-covalent assembly of peptide and DNA for viral vector encapsulation and cancer therapy.

Self-assembled amphiphilic peptide units and supercoiled, circular double-stranded plasmid DNA are used as building blocks to form peptide/DNA fibers for virus encapsulation. Since the fiber formation process takes place under ambient conditions and is aqueous-based without the use of denaturing organic solvents, the bioactivity of viruses is well preserved.

Human embryonic stem cell-derived mesenchymal stem cells as cellular delivery vehicles for prodrug gene therapy of glioblastoma.

Mesenchymal stem cells (MSCs) possess tumor-tropic properties and consequently have been used to deliver therapeutic agents for cancer treatment. Their potential in cancer therapy highlights the need for a consistent and renewable source for the production of uniform human MSCs suitable for clinical applications. In this study, we seek to investigate whether human embryonic stem cells can be used as a cell source to fulfill this goal.

Cell size and growth rate are major determinants of replicative lifespan.

Yeast cells, like mammalian cells, enlarge steadily as they age. Unabated cell growth can promote cellular senescence; however, the significance of the relationship between size and cellular lifespan is not well understood. Herein, we report a genetic link between cell size, growth rate and lifespan.

Polyethylenimine coating to produce serum-resistant baculoviral vectors for in vivo gene delivery.

Recombinant baculoviral vectors efficiently transduce many types of mammalian cells. However, their in vivo applications are hampered by the sensitivity of the virus to complement-mediated inactivation. Based on our observation that the surface charge of baculovirus is negative at neutral pH, we developed a procedure to coat baculoviral vectors with positively charged polyethylenimine 25 kDa, a commonly tested non-viral gene delivery vector, through electrostatic interaction.