Prompt Tuning in Biomedical Relation Extraction

Biomedical relation extraction (RE) is critical in constructing high-quality knowledge graphs and databases as well as supporting many downstream text mining applications. This paper explores prompt tuning on biomedical RE and its few-shot scenarios, aiming to propose a simple yet effective model for this specific task. Prompt tuning reformulates natural language processing (NLP) downstream tasks into masked language problems by embedding specific text prompts into the original input, facilitating the adaption of pre-trained language models (PLMs) to better address these tasks. This study presents a customized prompt tuning model designed explicitly for biomedical RE, including its applicability in few-shot learning contexts. The model's performance was rigorously assessed using the chemical-protein relation (CHEMPROT) dataset from BioCreative VI and the drug-drug interaction (DDI) dataset from SemEval-2013, showcasing its superior performance over conventional fine-tuned PLMs across both datasets, encompassing few-shot scenarios. This observation underscores the effectiveness of prompt tuning in enhancing the capabilities of conventional PLMs, though the extent of enhancement may vary by specific model. Additionally, the model demonstrated a harmonious balance between simplicity and efficiency, matching state-of-the-art performance without needing external knowledge or extra computational resources. The pivotal contribution of our study is the development of a suitably designed prompt tuning model, highlighting prompt tuning's effectiveness in biomedical RE. It offers a robust, efficient approach to the field's challenges and represents a significant advancement in extracting complex relations from biomedical texts.

Supplementary Information

The online version contains supplementary material available at 10.1007/s41666-024-00162-9.

Amanitin-induced variable cytotoxicity in various cell lines is mediated by the different expression levels of OATP1B3

Amanita phalloides is one of the deadliest mushrooms worldwide, causing most fatal cases of mushroom poisoning. Among the poisonous substances of Amanita phalloides, amanitins are the most lethal toxins to humans. Currently, there are no specific antidotes available for managing amanitin poisoning and treatments are lack of efficacy. Amanitin mainly causes severe injuries to specific organs, such as the liver, stomach, and kidney, whereas the lung, heart, and brain are hardly affected. However, the molecular mechanism of this phenomenon remains not understood. To explore the possible mechanism of organ specificity of amanitin-induced toxicity, eight human cell lines derived from different organs were exposed to α, β, and γ-amanitin at concentrations ranging from 0.3 to 100 μM. We found that the cytotoxicity of amanitin differs greatly in various cell lines, among which liver-derived HepG2, stomach-derived BGC-823, and kidney-derived HEK-293 cells are most sensitive. Further mechanistic study revealed that the variable cytotoxicity is mainly dependent on the different expression levels of the organic anion transporting polypeptide 1B3 (OATP1B3), which facilitates the internalization of amanitin into cells. Besides, knockdown of OATP1B3 in HepG2 cells prevented α-amanitin-induced cytotoxicity. These results indicated that OATP1B3 may be a crucial therapeutic target against amanitin-induced organ failure.

Identifying and Extracting Rare Diseases and Their Phenotypes with Large Language Models

Purpose

Phenotyping is critical for informing rare disease diagnosis and treatment, but disease phenotypes are often embedded in unstructured text. While natural language processing (NLP) can automate extraction, a major bottleneck is developing annotated corpora. Recently, prompt learning with large language models (LLMs) has been shown to lead to generalizable results without any (zero-shot) or few annotated samples (few-shot), but none have explored this for rare diseases. Our work is the first to study prompt learning for identifying and extracting rare disease phenotypes in the zero- and few-shot settings.

Methods

We compared the performance of prompt learning with ChatGPT and fine-tuning with BioClinicalBERT. We engineered novel prompts for ChatGPT to identify and extract rare diseases and their phenotypes (e.g., diseases, symptoms, and signs), established a benchmark for evaluating its performance, and conducted an in-depth error analysis.

Results

Overall, fine-tuning BioClinicalBERT resulted in higher performance (F1 of 0.689) than ChatGPT (F1 of 0.472 and 0.610 in the zero- and few-shot settings, respectively). However, ChatGPT achieved higher accuracy for rare diseases and signs in the one-shot setting (F1 of 0.778 and 0.725). Conversational, sentence-based prompts generally achieved higher accuracy than structured lists.

Conclusion

Prompt learning using ChatGPT has the potential to match or outperform fine-tuning BioClinicalBERT at extracting rare diseases and signs with just one annotated sample. Given its accessibility, ChatGPT could be leveraged to extract these entities without relying on a large, annotated corpus. While LLMs can support rare disease phenotyping, researchers should critically evaluate model outputs to ensure phenotyping accuracy.

Efficacy of intestinal microorganisms on immunotherapy of non-small cell lung cancer

While the 5-year survival rate of patients with advanced non-small cell lung cancer (NSCLC) has seen some improvement, the majority of NSCLC patients fail to respond to immunotherapy with immune checkpoint inhibitors (ICIs). It is critical to identify effective biomarkers that can enhance the efficacy of immunotherapy. The clinical data in the current study were collected from NSCLC patients treated with ICIs, and two groups were classified according to treatment effect: good group with consistent efficacy, poor group with only progressiveness. Differences in intestinal microbiota between the two groups were analyzed using 16s rRNA sequencing. Beta diversity analysis indicated differences between the two groups that were available for differentiation. Comparison of the number of common or unique operational taxonomic units (OTUs) among different groups suggested that there were 53 unique OTUs in the good group and 51 unique OTUs in the poor group. At the phylum level, there was a difference between the two groups for several bacterial groups with the highest abundance values, among which Firmicutes, Actinobacteria and Fusobacteria were more abundant in the good group. Members of the genera Bifidobacterium and Lactobacillus were abundant in the good group, while the abundance of Bacteroides was low. Biomarkers in the poor group included Bacteroides, Bacteroidetes, Bacteroidia, Bacteroidales, Bacteroidaceae and Veillonellaceae. The intestinal microbiota composition affected the immunotherapy process for NSCLC, which might offer more rational instructions for the clinical application of ICIs in NSCLC patients.

CDC73 serves as a tumour-promoting factor in oesophageal cancer

The role of human cell division cycle 73 (CDC73) in human cancers has sparked controversy; however, its significance in oesophageal cancer remains elusive. This study aimed to elucidate CDC73 expression and its biological implications in human oesophageal cancer. Our findings unveiled a notable upregulation of CDC73 in both oesophageal cancer cell lines and tissues. Importantly, elevated CDC73 levels in patients with oesophageal cancer correlated with an unfavourable prognosis. Functional investigations revealed that CDC73 knockdown effectively curtailed the proliferation and growth of oesophageal cancer cells both in vitro and in vivo. Mechanistically, RRP15 emerged as a potential downstream target of CDC73 through a screening process involving identification of the top co-expressed genes, subsequent knockdown experiments, and observation of significant inhibition of cell proliferation, with RRP15 showing the most pronounced effect. This finding was further supported by the positive correlation observed between CDC73 and RRP15 in ESCA samples analysed using the ENCORI Pan-Cancer Analysis Platform. Notably, depletion of RRP15 in CDC73-overexpressing cells led to a shift from augmented to diminished tumour growth. Collectively, our findings underscore the pivotal role of CDC73 in oesophageal cancer through the modulation of RRP15 expression, suggesting CDC73 as a potential therapeutic target for treating oesophageal cancer.

Abundance of Modifications in Mature miRNAs Revealed by LC-MS/MS Method Coupled with a Two-Step Hybridization Purification Strategy

Accurate detection of endogenous miRNA modifications, such as N6-methyladenosine (m6A), 7-methylguanosine (m7G), and 5-methylcytidine (m5C), poses significant challenges, resulting in considerable uncertainty regarding their presence in mature miRNAs. In this study, we demonstrate for the first time that liquid chromatography coupled with a tandem mass spectrometry (LC-MS/MS) nucleoside analysis method is a practical tool for quantitatively analyzing human miRNA modifications. The newly designed liquid-solid two-step hybridization (LSTH) strategy enhances specificity for miRNA purification, while LC-MS/MS offers robust capability in recognizing modifications and sufficient sensitivity with detection limits ranging from attomoles to low femtomoles. Therefore, it provides a more reliable approach compared to existing techniques for revealing modifications in endogenous miRNAs. With this approach, we characterized m6A, m7G, and m5C modifications in miR-21-5p, Let-7a/e-5p, and miR-10a-5p isolated from cultured cells and observed unexpectedly low abundance (<1% at each site) of these modifications.

IGF2BP2 regulates the inflammation of fibroblast-like synoviocytes via GSTM5 in rheumatoid arthritis

Rheumatoid arthritis (RA) is a chronic autoimmune disease with an unknown etiology. RA cannot be fully cured and requires lengthy treatment, imposing a significant burden on both individuals and society. Due to the lack of specific drugs available for treating RA, exploring a key new therapeutic target for RA is currently an important task. Activated fibroblast-like synoviocytes (FLSs) play a crucial role in the progression of RA, which release interleukin (IL)-1β, IL-6 and tumor necrosis factor (TNF)-α resulting in abnormal inflammatory reaction in the synovium. A previous study has highlighted the correlation of m6A reader insulin-like growth factor 2 mRNA-binding protein 2 (IGF2BP2) with inflammation-related diseases in human. However, the role of IGF2BP2 in the inflammatory reaction of FLSs during RA progression has not been assessed. In this study, IGF2BP2 expression was decreased in the synovial tissues of RA patients and collagen-induced arthritis (CIA) rats. Intra-articular injection of an adeno-associated virus (AAV) vector overexpressing IGF2BP2 relieved paw swelling, synovial hyperplasia and cartilage destruction in CIA rats. IGF2BP2 overexpression also inhibited lipopolysaccharide (LPS)-mediated RA fibroblast-like synoviocytes (RA-FLSs) migration and invasion accompanied by a decreased level of inflammatory factors in vitro. Conversely, IGF2BP2 suppression promoted RA-FLSs migration and invasion with an elevated level of inflammatory factors in vitro. The sequencing result showed that glutathione S-transferase Mu 5 (GSTM5), a key antioxidant gene, was the target mRNA of IGF2BP2. Further experiments demonstrated that IGF2BP2 strengthened the stability of GSTM5 mRNA, leading to weakened inflammatory reaction and reduced expression of matrix metalloproteinase 9 and 13 (MMP9, MMP13). Therefore, IGF2BP2-GSTM5 axis may represent a potential therapeutic target for RA treatment.

Mechanically Tunable Transmittance Convection Shield for Dynamic Radiative Cooling

Radiative cooling is the process to dissipate heat to the outer space through an atmospheric window (8-13 μm), which has great potential for energy savings in buildings. However, the traditional "static" spectral characteristics of radiative cooling materials may result in overcooling during the cold season or at night, necessitating the development of dynamic spectral radiative cooling for enhanced energy saving potential. In this study, we showcase the realization of dynamic radiative cooling by modulating the heat transfer process using a tunable transmittance convection shield (TTCS). The transmittance of the TTCS in both solar spectrum and atmospheric window can be dynamically adjusted within ranges of 28.8-72.9 and 27.0-80.5%, with modulation capabilities of ΔTsolar = 44.1% and ΔT8-13 μm = 53.5%, respectively. Field measurements demonstrate that through the modulation, the steady-state temperature of the TTCS architecture is 0.3 °C lower than that of a traditional radiative cooling architecture during the daytime and 3.3 °C higher at nighttime, indicating that the modulation strategy can effectively address the overcooling issue, offering an efficient way of energy saving through dynamic radiative cooling.

Gate Voltage- and Bias Voltage-Tunable Staggered-Gap to Broken-Gap Transition Based on WSe2/Ta2NiSe5 Heterostructure for Multimode Optoelectronic Logic Gate

Two-dimensional (2D) materials with superior properties exhibit tremendous potential in developing next-generation electronic and optoelectronic devices. Integrating various functions into one device is highly expected as that endows 2D materials great promise for more Moore and more-than-Moore device applications. Here, we construct a WSe2/Ta2NiSe5 heterostructure by stacking the p-type WSe2 and the n-type narrow gap Ta2NiSe5 with the aim to achieve a multifunction optoelectronic device. Owing to the large interface potential barrier, the heterostructure device reveals a prominent diode feature with a large rectify ratio (7.6 × 104) and a low dark current (10-12 A). Especially, gate voltage- and bias voltage-tunable staggered-gap to broken-gap transition is achieved on the heterostructure device, which enables gate voltage-tunable forward and reverse rectifying features. As results, the heterostructure device exhibits superior self-powered photodetection properties, including a high detectivity of 1.08 × 1010 Jones and a fast response time of 91 μs. Additionally, the intrinsic structural anisotropy of Ta2NiSe5 endows the heterostructure device with strong polarization-sensitive photodetection and high-resolution polarization imaging. Based on these characteristics, a multimode optoelectronic logic gate is realized on the heterostructure via synergistically modulating the light on/off, polarization angle, gate voltage, and bias voltage. This work shed light on the future development of constructing high-performance multifunctional optoelectronic devices.

[Clinical efficacy of intraarticular vancomycin in preventing early periprosthetic joint infection after primary knee arthroplasty]

Objective: To investigate the clinical effect of intraarticular vancomycin on early periprosthetic joint infection (PJI) in knee arthroplasty and the incidence of postoperative complications. Methods: This is a retrospective cohort study. The clinical data of 1 867 patients who underwent primary knee arthroplasty at Department of Joint Surgery, the Affiliated Hospital of Qingdao University from April 2022 to June 2023 were retrospectively analysed, including total knee arthroplasty (TKA), robotic-assisted total knee arthroplasty (RA-TKA) and unicondylar knee arthroplasty (UKA). There were 687 males and 1180 females, aged (68.0±11.2)years(range:45 to 87 years). Patients were divided into the vancomycin group and the control group according to whether or not intra-articular injection of 1 g of vancomycin powder dissolved in 30 ml of saline was performed after intraoperative joint capsule closure. In the vancomycin group, 925 patients were included, including 782 TKA, 27 RA-TKA and 116 UKA.In the control group, 942 patients were included, including 767 TKA, 99 RA-TKA and 76 UKA. Early PJI, wound complications, and vancomycin-related toxicity including acute renal collapse, ototoxicity, and allergic reactions were assessed within 3 months postoperatively. Results: No PJI was found in all patients in the vancomycin group.Five cases (0.7%,5/767) of early PJI were found in TKA patients in the control group, with a statistically significant difference (P=0.030); 1 case of early PJI was found in each RA-TKA and UKA patients, with non-significant difference compared with vancomycin group (all P>0.05). Two cases (0.3%,2/782) of incisional complications were found in TKA patients in the vancomycin group, and 4 cases (0.5%, 4/767) of incisional complications were found in TKA patients in the control group, with non-significant difference(P=0.449); no incisional complications were found in RA-TKA patients in the vancomycin group, and 1 case (1.0%,1/99) of incisional complications were found in RA-TKA patients in the control group, the difference was not statistically significant (P>0.05); no incisional complications were found in both groups of UKA patients.No vancomycin-related acute kidney injury, ototoxicity, or allergic reactions was observed in all patients. Conclusion: Intra-articular injection of 1 g of vancomycin suspension after arthrotomy closure during TKA maybe lower the risk of early PJI without increasing the risk of wound complication and vancomycin-associated systemic toxicity.

Process management program to prevent falls in hospitalized patients with neuropsychiatric disorders: a quality improvement program

BACKGROUND

Falls were among the most common adverse nursing events. The incidence of falls in patients with neuropsychiatric disorders was high, and the occurrence of falls not only caused physical and psychological harm to patients but also led to medical disputes. Therefore, interventions for falls prevention were essential, but evaluations of the intervention process were lacking.

METHODS

In this study, a process management program to prevent falls based on the "structure-process-outcome" quality evaluation model was designed and applied to the clinical practice of falls prevention in hospitalized patients with neuropsychiatric disorders. The process quality evaluation checklist to prevent falls was used to supervise the implementation effect of intervention measures to prevent falls, identify the problems in the intervention measures, and make continuous improvements, to reduce the incidence of falls in such hospitalized patients as the final index.

RESULTS

The incidence of inpatient falls decreased from 0.199‰ (0.199 per 1000 patient-days) to 0.101‰ (0.101 per 1000 patient-days) before and after the implementation of the process management program for 12 months, 24 months, and 36 months, respectively, and the difference was statistically significant (P < .05). The probability of falls was reduced by 49% after 36 months of monitoring. Furthermore, the proportion of patients at high risk of falls exhibited a downward trend.

CONCLUSION

This quality improvement program was feasible and effective at reducing falls in hospitalized patients with neuropsychiatric disorders. Therefore, attention should be given to monitoring process quality in the management of falls.

2D Free-Standing GeS1-xSex with Composition-Tunable Bandgap for Tailored Polarimetric Optoelectronics

Germanium-based monochalcogenides (i.e., GeS and GeSe) with desirable properties are promising candidates for the development of next-generation optoelectronic devices. However, they are still stuck with challenges, such as relatively fixed electronic band structure, unconfigurable optoelectronic characteristics, and difficulty in achieving free-standing growth. Herein, it is demonstrated that two-dimensional (2D) free-standing GeS1-xSex (0 ≤ x ≤ 1) nanoplates can be grown by low-pressure rapid physical vapor deposition (LPRPVD), fulfilling a continuously composition-tunable optical bandgap and electronic band structure. By leveraging the synergistic effect of composition-dependent modulation and free-standing growth, GeS1-xSex-based optoelectronic devices exhibit significantly configurable hole mobility from 6.22 × 10-4 to 1.24 cm2V-1s⁻1 and tunable responsivity from 8.6 to 311 A W-1 (635 nm), as x varies from 0 to 1. Furthermore, the polarimetric sensitivity can be tailored from 4.3 (GeS0.29Se0.71) to 1.8 (GeSe) benefiting from alloy engineering. Finally, the tailored imaging capability is also demonstrated to show the application potential of GeS1-xSex alloy nanoplates. This work broadens the functionality of conventional binary materials and motivates the development of tailored polarimetric optoelectronic devices.

Large language models for biomedicine: foundations, opportunities, challenges, and best practices

OBJECTIVES

Generative large language models (LLMs) are a subset of transformers-based neural network architecture models. LLMs have successfully leveraged a combination of an increased number of parameters, improvements in computational efficiency, and large pre-training datasets to perform a wide spectrum of natural language processing (NLP) tasks. Using a few examples (few-shot) or no examples (zero-shot) for prompt-tuning has enabled LLMs to achieve state-of-the-art performance in a broad range of NLP applications. This article by the American Medical Informatics Association (AMIA) NLP Working Group characterizes the opportunities, challenges, and best practices for our community to leverage and advance the integration of LLMs in downstream NLP applications effectively. This can be accomplished through a variety of approaches, including augmented prompting, instruction prompt tuning, and reinforcement learning from human feedback (RLHF).

TARGET AUDIENCE

Our focus is on making LLMs accessible to the broader biomedical informatics community, including clinicians and researchers who may be unfamiliar with NLP. Additionally, NLP practitioners may gain insight from the described best practices.

SCOPE

We focus on 3 broad categories of NLP tasks, namely natural language understanding, natural language inferencing, and natural language generation. We review the emerging trends in prompt tuning, instruction fine-tuning, and evaluation metrics used for LLMs while drawing attention to several issues that impact biomedical NLP applications, including falsehoods in generated text (confabulation/hallucinations), toxicity, and dataset contamination leading to overfitting. We also review potential approaches to address some of these current challenges in LLMs, such as chain of thought prompting, and the phenomena of emergent capabilities observed in LLMs that can be leveraged to address complex NLP challenge in biomedical applications.

[Establishment and application of a database for hereditary kidney disease in Chinese children]

Hereditary kidney diseases are common causes of chronic kidney disease (CKD) in children and adolescents, and also has an important role in the onset and progression of CKD in adulthood. Constructing a data sharing registration system for hereditary kidney disease and forming representative data with Chinese population specificity, is of great significance for achieving phenotype and genotype characterization, improving precision management level and mechanism research. The high heterogeneity of the disease and the scattered distribution of patients have led to a lack of understanding and unified management standards for hereditary kidney disease. Led by pediatric nephrology specialists and geneticists, integrating data sources from various centers can leverage clinical resource advantages. Focusing on different subtype disease cohorts, integrating and analyzing data such as genotype, multi-omics, and clinical outcomes, can achieve breakthroughs in the key points of disease diagnosis and treatment.

Culture-free detection of β-lactamase-Producing bacteria in urinary tract infections using a paper sensor

Developing simple, inexpensive, fast, sensitive, and specific probes for antibiotic-resistant bacteria is crucial for the management of urinary tract infections (UTIs). We here propose a paper-based sensor for the rapid detection of β-lactamase-producing bacteria in the urine samples of UTI patients. By conjugating a strongly electronegative group -N+(CH3)3 with the core structures of cephalosporin and carbapenem antibiotics, two visual probes were achieved to respectively target the extended-spectrum/AmpC β-lactamases (ESBL/AmpC) and carbapenemase, the two most prevalent factors causing antibiotic resistance. By integrating these probes into a portable paper sensor, we confirmed 10 and 8 cases out of 30 clinical urine samples as ESBL/AmpC- and carbapenemase-positive, respectively, demonstrating 100% clinical sensitivity and specificity. This paper sensor can be easily conducted on-site, without resorting to bacterial culture, providing a solution to the challenge of rapid detection of β-lactamase-producing bacteria, particularly in resource-limited settings.

A span-based model for extracting overlapping PICO entities from randomized controlled trial publications

OBJECTIVES

Extracting PICO (Populations, Interventions, Comparison, and Outcomes) entities is fundamental to evidence retrieval. We present a novel method, PICOX, to extract overlapping PICO entities.

MATERIALS AND METHODS

PICOX first identifies entities by assessing whether a word marks the beginning or conclusion of an entity. Then, it uses a multi-label classifier to assign one or more PICO labels to a span candidate. PICOX was evaluated using 1 of the best-performing baselines, EBM-NLP, and 3 more datasets, ie, PICO-Corpus and randomized controlled trial publications on Alzheimer's Disease (AD) or COVID-19, using entity-level precision, recall, and F1 scores.

RESULTS

PICOX achieved superior precision, recall, and F1 scores across the board, with the micro F1 score improving from 45.05 to 50.87 (P ≪.01). On the PICO-Corpus, PICOX obtained higher recall and F1 scores than the baseline and improved the micro recall score from 56.66 to 67.33. On the COVID-19 dataset, PICOX also outperformed the baseline and improved the micro F1 score from 77.10 to 80.32. On the AD dataset, PICOX demonstrated comparable F1 scores with higher precision when compared to the baseline.

CONCLUSION

PICOX excels in identifying overlapping entities and consistently surpasses a leading baseline across multiple datasets. Ablation studies reveal that its data augmentation strategy effectively minimizes false positives and improves precision.

The Enhanced Performance of Oxide Thin-Film Transistors Fabricated by a Two-Step Deposition Pressure Process

It is usually difficult to realize high mobility together with a low threshold voltage and good stability for amorphous oxide thin-film transistors (TFTs). In addition, a low fabrication temperature is preferred in terms of enhancing compatibility with the back end of line of the device. In this study, α-IGZO TFTs were prepared by high-power impulse magnetron sputtering (HiPIMS) at room temperature. The channel was prepared under a two-step deposition pressure process to modulate its electrical properties. X-ray photoelectron spectra revealed that the front-channel has a lower Ga content and a higher oxygen vacancy concentration than the back-channel. This process has the advantage of balancing high mobility and a low threshold voltage of the TFT when compared with a conventional homogeneous channel. It also has a simpler fabrication process than that of a dual active layer comprising heterogeneous materials. The HiPIMS process has the advantage of being a low temperature process for oxide TFTs.

Triple-negative breast cancer survival prediction using artificial intelligence through integrated analysis of tertiary lymphoid structures and tumor budding

BACKGROUND

Triple-negative breast cancer (TNBC) is a highly heterogeneous and clinically aggressive disease. Accumulating evidence indicates that tertiary lymphoid structures (TLSs) and tumor budding (TB) are significantly correlated with the outcomes of patients who have TNBC, but no integrated TLS-TB profile has been established to predict their survival. The objective of this study was to investigate the relationship between the TLS/TB ratio and clinical outcomes of patients with TNBC using artificial intelligence (AI)-based analysis.

METHODS

The infiltration levels of TLSs and TB were evaluated using hematoxylin and eosin staining, immunohistochemistry staining, and AI-based analysis. Various cellular subtypes within TLS were determined by multiplex immunofluorescence. Subsequently, the authors established a nomogram model, conducted calibration curve analyses, and performed decision curve analyses using R software.

RESULTS

In both the training and validation cohorts, the antitumor/protumor model established by the authors demonstrated a positive correlation between the TLS/TB index and the overall survival (OS) and relapse-free survival (RFS) of patients with TNBC. Notably, patients who had a high percentage of CD8-positive T cells, CD45RO-positive T cells, or CD20-positive B cells within the TLSs experienced improved OS and RFS. Furthermore, the authors developed a comprehensive TLS-TB profile nomogram based on the TLS/TB index. This novel model outperformed the classical tumor-lymph node-metastasis staging system in predicting the OS and RFS of patients with TNBC.

CONCLUSIONS

A novel strategy for predicting the prognosis of patients with TNBC was established through integrated AI-based analysis and a machine-learning workflow. The TLS/TB index was identified as an independent prognostic factor for TNBC. This nomogram-based TLS-TB profile would help improve the accuracy of predicting the prognosis of patients who have TNBC.

Repurposing non-pharmacological interventions for Alzheimer's disease through link prediction on biomedical literature

Non-pharmaceutical interventions (NPI) have great potential to improve cognitive function but limited investigation to discover NPI repurposing for Alzheimer's Disease (AD). This is the first study to develop an innovative framework to extract and represent NPI information from biomedical literature in a knowledge graph (KG), and train link prediction models to repurpose novel NPIs for AD prevention. We constructed a comprehensive KG, called ADInt, by extracting NPI information from biomedical literature. We used the previously-created SuppKG and NPI lexicon to identify NPI entities. Four KG embedding models (i.e., TransE, RotatE, DistMult and ComplEX) and two novel graph convolutional network models (i.e., R-GCN and CompGCN) were trained and compared to learn the representation of ADInt. Models were evaluated and compared on two test sets (time slice and clinical trial ground truth) and the best performing model was used to predict novel NPIs for AD. Discovery patterns were applied to generate mechanistic pathways for high scoring candidates. The ADInt has 162,212 nodes and 1,017,284 edges. R-GCN performed best in time slice (MR = 5.2054, Hits@10 = 0.8496) and clinical trial ground truth (MR = 3.4996, Hits@10 = 0.9192) test sets. After evaluation by domain experts, 10 novel dietary supplements and 10 complementary and integrative health were proposed from the score table calculated by R-GCN. Among proposed novel NPIs, we found plausible mechanistic pathways for photodynamic therapy and Choerospondias axillaris to prevent AD, and validated psychotherapy and manual therapy techniques using real-world data analysis. The proposed framework shows potential for discovering new NPIs for AD prevention and understanding their mechanistic pathways.

A Principal Component Regression-Based Electrophysiological Study of Patients with Severe Infections

Objective

Severe infections can lead to neuromyopathy in critically ill patients, resulting in limb weakness and difficulty in weaning from a ventilator. This study aims to assess the electrophysiological test results in patients with severe infection and their correlation with severity scores (APACHE II and SOFA).

Methods

Thirty-one patients with severe infection in the EICU were prospectively studied. Factor analysis and principal component regression were applied to develop linear models of electrophysiological diagnostic outcomes with APACHE II and SOFA scores for the entire patient cohort, the younger group (age<55) cohort, and the older group (age>55) cohort of patients with severe infections, respectively.

Results

Among patients with a severe infection in the EICU, the proportion of patients without critical neuromyopathy with more than 50% F-wave presence in the median, ulnar, and tibial nerves (64.9%, 56.8%, 48.6%, respectively) was significantly higher than in the group with critical neuromyopathy (52.1%, 35.4%, 29.2%, respectively.), and the proportion of patients with critical neuromyopathy who did not elicit the three types of F wave was significantly higher in the cohort of patients with critical neuromyopathy (40.5%, 32.4%, 35.1%, respectively.) were significantly higher than in the cohort of patients without critical illness (18.8%, 12.5%, 20.8%, respectively). In addition, on average, patients with critical neuromyopathy had a much lower CMAP for the median nerve (wrist, elbow) (2.4, 1.88, respectively) (4.3, 3.9, respectively in undiagnosed cohort), ulnar nerve (wrist, elbow) (2.4, 1.88, respectively) (5.65, 5.4, respectively in undiagnosed cohort), and tibial nerve(ankle, popliteal fossa) (2.7, 1.57, respectively)(6.55, 5.3, respectively in undiagnosed cohort) nerves than patients without critical neuromyopathy, and showed more non-elicitation, which was not seen in the cohort of patients without critical neuromyopathy. The CMAP returned to normal in the cohort of patients without critical neuromyopathy. Therefore, with respect to our selected electrophysiological parameters, the two patient groups showed significant differences in terms of the specific values and statistical analysis (Table 1). Through factor analysis and principal component regression, we found that CMAP and F-wave were highly correlated with APACHE II and SOFA scores, and the correlation between the electrophysiological wave spectrum and the two scores was further quantified by principal component regression.

Conclusion

Electrophysiological spectroscopy can serve as an early warning for the development of neuromuscular disease in EICU patients. Abnormal electrophysiological diagnosis prior to actual neuromuscular abnormalities and its subsequent return to normal can help identify high-risk patients and implement early interventions.

Confidence score: a data-driven measure for inclusive systematic reviews considering unpublished preprints

OBJECTIVES

COVID-19, since its emergence in December 2019, has globally impacted research. Over 360 000 COVID-19-related manuscripts have been published on PubMed and preprint servers like medRxiv and bioRxiv, with preprints comprising about 15% of all manuscripts. Yet, the role and impact of preprints on COVID-19 research and evidence synthesis remain uncertain.

MATERIALS AND METHODS

We propose a novel data-driven method for assigning weights to individual preprints in systematic reviews and meta-analyses. This weight termed the "confidence score" is obtained using the survival cure model, also known as the survival mixture model, which takes into account the time elapsed between posting and publication of a preprint, as well as metadata such as the number of first 2-week citations, sample size, and study type.

RESULTS

Using 146 preprints on COVID-19 therapeutics posted from the beginning of the pandemic through April 30, 2021, we validated the confidence scores, showing an area under the curve of 0.95 (95% CI, 0.92-0.98). Through a use case on the effectiveness of hydroxychloroquine, we demonstrated how these scores can be incorporated practically into meta-analyses to properly weigh preprints.

DISCUSSION

It is important to note that our method does not aim to replace existing measures of study quality but rather serves as a supplementary measure that overcomes some limitations of current approaches.

CONCLUSION

Our proposed confidence score has the potential to improve systematic reviews of evidence related to COVID-19 and other clinical conditions by providing a data-driven approach to including unpublished manuscripts.

[Influence of paraspinal muscle degeneration and postoperative Roussouly classification restoration on mechanical complications in female patients with degenerative scoliosis after surgery]

Objective: To investigate the impact of lumbar paraspinal muscle degeneration and postoperative failure to restore ideal Roussouly classification on the occurrence of mechanical complications (MC) following long-segment spinal correction surgery in female patients with degenerative scoliosis (DS). Methods: The clinical data of 72 female DS patients who underwent long-segment spinal correction surgery in Gulou Hospital from June 2017 to November 2021 were retrospectively analyzed. According to whether restoring the ideal Roussouly classification after surgery, the patients were divided into R group(recovery group) (n=51) and N group(non-recovery group) (n=21). According to whether mechanical complications occurred after operation within two years, the patients were divided into MC (mechanical complications)group (n=24) and NMC(non-mechanical complications) group (n=48). The RM group (n=14) experienced mechanical complications in the R group, while the RN group (n=37) did not. The NM group (n=10) experienced mechanical complications in the N group, while the NN group (n=11) did not.Radiographic assessment included Sagittal parameters of spine and pelvis, standardized cross-sectional area (SCSA) and fat infiltration rate (FI%) of paraspinal muscle at each lumbar disc level. Results: The age of DS patients in this study was (61.4±6.2) years.The incidence of MC was 33.33%(n=24)in all patients. The incidence of MC was 27.45%(n=14)in group R and 47.62%(n=10) in group N. The correction amount of pelvic tilt angle (PT) (-11.62°±10.06° vs -7.04°±8.45°, P=0.046) and T1 pelvic angle(TPA)(-12.88°±11.23° vs -7.31°±9.55°, P=0.031)during surgery were significantly higher in MC group compared to the NMC group. In group R, the FI% of paraspinal muscles in each lumbar segment of patients with postoperative MC was higher than that in patients without MC (P<0.05). In the R and N groups, there was no significant difference inthe SCSA of the lumbar paravertebral muscles between patients with postoperative MC and those without MC at each level (all P>0.05). Multivariate logistic regression analysis showed that the average FI% of lumbar PSM was correlated with the occurrence of MC after spinal fusion in DS patients.The average FI% of lumbar PSM≥22.63% was a risk factors for MC after spinal fusion (P=0.010,OR=1.088, 95%CI:1.020-1.160). Conclusions: Female DS patients with higher degree of preoperative paraspinal muscle degeneration have a higher incidence of postoperative mechanical complications. For these patients,.there is still a higher risk of mechanical complications after surgery even if the ideal Roussouly classification is restored after surgery.

Artificial intelligence-powered pharmacovigilance: A review of machine and deep learning in clinical text-based adverse drug event detection for benchmark datasets

OBJECTIVE

The primary objective of this review is to investigate the effectiveness of machine learning and deep learning methodologies in the context of extracting adverse drug events (ADEs) from clinical benchmark datasets. We conduct an in-depth analysis, aiming to compare the merits and drawbacks of both machine learning and deep learning techniques, particularly within the framework of named-entity recognition (NER) and relation classification (RC) tasks related to ADE extraction. Additionally, our focus extends to the examination of specific features and their impact on the overall performance of these methodologies. In a broader perspective, our research extends to ADE extraction from various sources, including biomedical literature, social media data, and drug labels, removing the limitation to exclusively machine learning or deep learning methods.

METHODS

We conducted an extensive literature review on PubMed using the query "(((machine learning [Medical Subject Headings (MeSH) Terms]) OR (deep learning [MeSH Terms])) AND (adverse drug event [MeSH Terms])) AND (extraction)", and supplemented this with a snowballing approach to review 275 references sourced from retrieved articles.

RESULTS

In our analysis, we included twelve articles for review. For the NER task, deep learning models outperformed machine learning models. In the RC task, gradient Boosting, multilayer perceptron and random forest models excelled. The Bidirectional Encoder Representations from Transformers (BERT) model consistently achieved the best performance in the end-to-end task. Future efforts in the end-to-end task should prioritize improving NER accuracy, especially for 'ADE' and 'Reason'.

CONCLUSION

These findings hold significant implications for advancing the field of ADE extraction and pharmacovigilance, ultimately contributing to improved drug safety monitoring and healthcare outcomes.

FedFSA: Hybrid and federated framework for functional status ascertainment across institutions

INTRODUCTION

Patients' functional status assesses their independence in performing activities of daily living, including basic ADLs (bADL), and more complex instrumental activities (iADL). Existing studies have discovered that patients' functional status is a strong predictor of health outcomes, particularly in older adults. Depite their usefulness, much of the functional status information is stored in electronic health records (EHRs) in either semi-structured or free text formats. This indicates the pressing need to leverage computational approaches such as natural language processing (NLP) to accelerate the curation of functional status information. In this study, we introduced FedFSA, a hybrid and federated NLP framework designed to extract functional status information from EHRs across multiple healthcare institutions.

METHODS

FedFSA consists of four major components: 1) individual sites (clients) with their private local data, 2) a rule-based information extraction (IE) framework for ADL extraction, 3) a BERT model for functional status impairment classification, and 4) a concept normalizer. The framework was implemented using the OHNLP Backbone for rule-based IE and open-source Flower and PyTorch library for federated BERT components. For gold standard data generation, we carried out corpus annotation to identify functional status-related expressions based on ICF definitions. Four healthcare institutions were included in the study. To assess FedFSA, we evaluated the performance of category- and institution-specific ADL extraction across different experimental designs.

RESULTS

ADL extraction performance ranges from an F1-score of 0.907 to 0.986 for bADL and 0.825 to 0.951 for iADL across the four healthcare sites. The performance for ADL extraction with impairment ranges from an F1-score of 0.722 to 0.954 for bADL and 0.674 to 0.813 for iADL across four healthcare sites. For category-specific ADL extraction, laundry and transferring yielded relatively high performance, while dressing, medication, bathing, and continence achieved moderate-high performance. Conversely, food preparation and toileting showed low performance.

CONCLUSION

NLP performance varied across ADL categories and healthcare sites. Federated learning using a FedFSA framework performed higher than non-federated learning for impaired ADL extraction at all healthcare sites. Our study demonstrated the potential of the federated learning framework in functional status extraction and impairment classification in EHRs, exemplifying the importance of a large-scale, multi-institutional collaborative development effort.

Foliar application of glycine-functionalized nanopesticides for effective prevention and control of root-knot nematodes via a targeted delivery strategy

BACKGROUND

Root-knot nematodes (RKNs) are the highly damaging pests for various crops, and the prevalence of RKNs has posed serious threats to worldwide agricultural harvest, severely affecting global food security and ecosystem health. Traditional pesticide systems on controlling RKNs generally cause environmental hazards and phytotoxicity due to the excessive use of pesticides resulted from low utilization efficiency. And effective approaches with biosafe and efficient features are highly demanded to break away from the dilemma caused by RKNs.

RESULTS

In this research, a nanopesticide system with root-targeted delivery function was developed to achieve effective prevention and control of RKNs. The nanocarriers (MSN-KH560-Gly) and the obtained nanopesticides (EB@MSN-KH560-Gly) were proved to be biosafe. Also, this nanopesticide system demonstrated sustained release behavior. The grafting of glycine (Gly) significantly improved the pesticide contents translocating to cucumber roots (about 304.7%). Besides, such root-targeted delivery function resulted in no root nodule in cucumber plants after the foliar application of these nanopesticides (prevention rate of 100%). In addition, the root nodule numbers of the infected cucumber plants decreased by 71.67%.

CONCLUSION

Foliar application of these Gly-functionalized nanopesticides achieved effective prevention and control of RKNs due to the root-targeted delivery property inherent in this nanopesticide system, and such root-targeted delivery strategy opens a novel and efficient method to protect crops from RKN invasion and thus facilitates the development of sustainable agriculture. © 2023 Society of Chemical Industry.

Covalent hits and where to find them

Covalent hits for drug discovery campaigns are neither fantastic beasts nor mythical creatures, they can be routinely identified through electrophile-first screening campaigns using a suite of different techniques. These include biophysical and biochemical methods, cellular approaches, and DNA-encoded libraries. Employing best practice, however, is critical to success. The purpose of this review is to look at state of the art covalent hit identification, how to identify hits from a covalent library and how to select compounds for medicinal chemistry programmes.

Efficacy comparison of 21 interventions to prevent retinopathy of prematurity: a Bayesian network meta-analysis of randomized controlled trials

BACKGROUND

The objective of this study was to evaluate the comparative efficacy of current interventions for the prevention of retinopathy of prematurity (ROP) in premature infants.

METHODS

A network meta-analysis (NMA) was performed. We searched PubMed, Web of Science, Scopus, Embase, and the Cochrane Library for relevant studies from their inception to May 5, 2022. Publications were eligible for our study if they were randomized controlled trials (RCTs) involving preterm infants at <37 weeks of gestational age and reported the incidence of any-stage ROP treated with the interventions of interest. The overall effect was pooled using the random effects model.

RESULTS

We identified 106 RCTs (involving 23894 participants). This NMA showed that vitamin A supplementation markedly reduced the incidence of ROP, in comparison with placebo (odds ratio [OR] = 0.59, 95% credible interval [95% CrI] 0.33, 0.85), fish oil-based lipid emulsion (OR = 0.57, 95% CrI 0.24, 0.90), early erythropoietin (OR = 0.51, 95% CrI 0.34, 0.98), probiotics (OR = 0.48, 95% CrI 0.32, 0.97), and human milk (OR = 0.50, 95% CrI 0.21, 0.78). Vitamin A supplementation has the highest probability of being the best option for reducing the ROP risk compared with the other 20 interventions based on its surface under the cumulative ranking curve (SUCRA) value (SUCRA = 92.50%, 95% CrI 0.71, 1.00).

CONCLUSIONS

Our findings suggest that among 21 interventions, vitamin A supplementation might be the best method of preventing ROP. This NMA offers an important resource for further efforts to develop preventive strategies for ROP.

Social and Behavior Factors of Alzheimer's Disease and Related Dementias: A National Study in the U.S

INTRODUCTION

Considerable research has linked many risk factors to Alzheimer's Disease and Related Dementias (ADRD). Without a clear etiology of ADRD, it is advantageous to rank the known risk factors by their importance and determine if disparities exist. Statistical-based ranking can provide insight into which risk factors should be further evaluated.

METHODS

This observational, population-based study assessed 50 county-level measures and estimates related to ADRD in 3,155 counties in the U.S. using data from 2010 to 2021. Statistical analysis was performed in 2022-2023. The machine learning method, eXtreme Gradient Boosting, was utilized to rank the importance of these variables by their relative contribution to the model performance. Stratified ranking was also performed based on a county's level of disadvantage. Shapley Additive exPlanations (SHAP) provided marginal contributions for each variable.

RESULTS

The top three ranked predictors at the county level were insufficient sleep, consuming less than one serving of fruits/vegetables per day among adults, and having less than a high school diploma. In both disadvantaged and non-disadvantaged counties, demographic variables such as sex and race were important in predicting ADRD. Lifestyle factors ranked highly in non-disadvantaged counties compared to more environmental factors in disadvantaged counties.

CONCLUSIONS

This ranked list of factors can provide a guided approach to ADRD primary prevention strategies in the U.S., as the effects of sleep, diet, and education on ADRD can be further developed. While sleep, diet, and education are important nationally, differing prevention strategies could be employed based on a county's level of disadvantage.

A Deep Learning Approach for Automated Extraction of Functional Status and New York Heart Association Class for Heart Failure Patients During Clinical Encounters

Introduction

Serial functional status assessments are critical to heart failure (HF) management but are often described narratively in documentation, limiting their use in quality improvement or patient selection for clinical trials. We developed and validated a deep learning-based natural language processing (NLP) strategy to extract functional status assessments from unstructured clinical notes.

Methods

We identified 26,577 HF patients across outpatient services at Yale New Haven Hospital (YNHH), Greenwich Hospital (GH), and Northeast Medical Group (NMG) (mean age 76.1 years; 52.0% women). We used expert annotated notes from YNHH for model development/internal testing and from GH and NMG for external validation. The primary outcomes were NLP models to detect (a) explicit New York Heart Association (NYHA) classification, (b) HF symptoms during activity or rest, and (c) functional status assessment frequency.

Results

Among 3,000 expert-annotated notes, 13.6% mentioned NYHA class, and 26.5% described HF symptoms. The model to detect NYHA classes achieved a class-weighted AUROC of 0.99 (95% CI: 0.98-1.00) at YNHH, 0.98 (0.96-1.00) at NMG, and 0.98 (0.92-1.00) at GH. The activity-related HF symptom model achieved an AUROC of 0.94 (0.89-0.98) at YNHH, 0.94 (0.91-0.97) at NMG, and 0.95 (0.92-0.99) at GH. Deploying the NYHA model among 166,655 unannotated notes from YNHH identified 21,528 (12.9%) with NYHA mentions and 17,642 encounters (10.5%) classifiable into functional status groups based on activity-related symptoms.

Conclusions

We developed and validated an NLP approach to extract NYHA classification and activity-related HF symptoms from clinical notes, enhancing the ability to track optimal care and identify trial-eligible patients.

Sudomotor dysfunction is associated with impaired left ventricular diastolic function in persons with type 2 diabetes: a cross-sectional study

BACKGROUND

The incidence of preserved ejection fraction heart failure has significantly increased in persons with type 2 diabetes mellitus (T2DM). Left ventricular (LV) diastolic dysfunction is an early and important manifestation of preserved ejection fraction heart failure. The onset of heart failure in persons with diabetes is associated with diabetic neuropathy. However, the relationship among sudomotor function, which is an early manifestation of small fiber neuropathy, and LV diastolic function remains unclear. This study aimed to explore the association between sudomotor function and LV diastolic function in persons with T2DM.

METHODS

In total, 699 persons with T2DM were enrolled and divided into three groups according to electrochemical skin conductance (ESC) assessed using the SUDOSCAN device: "no dysfunction" group (NSF), "moderate dysfunction" group (MDF), and "severe dysfunction" group (SDF). LV diastolic function was assessed using Doppler echocardiography. To evaluate the relationship between ESC and echocardiographic parameters, Pearson's correlation analysis was performed. Additionally, logistic regression analysis was used to determine the association between LV diastolic function and ESC. A receiver operating characteristic (ROC) curve was constructed to evaluate the performance of sudomotor function indicators in detecting impaired cardiac diastolic function.

RESULTS

There were 301 persons (43.06%) in the NSF group, 232 (33.19%) in the MDF group, and 166 (23.75%) in the SDF group. Compared to the NSF group, the MDF and SDF groups had higher A and E/e' and lower e' values (all p < 0.05). Pearson's correlation analysis showed that A and E/e' were negatively associated with foot ESC (FESC) and hand ESC (HESC), whereas e' was positively associated with FESC and HESC (all p < 0.05). After adjusting for confounding factors, binary logistic regression analysis showed that ESC was independently associated with impaired LV diastolic function (p = 0.003). The area under the ROC curve values for FESC and HESC were 0.621 and 0.635, respectively (both p < 0.05).

CONCLUSIONS

Deteriorating sudomotor function was associated with reduced diastolic function indicators. ESC can be used as a biomarker for detecting LV diastolic impairment.

Developing deep learning-based strategies to predict the risk of hepatocellular carcinoma among patients with nonalcoholic fatty liver disease from electronic health records

OBJECTIVE

The accuracy of deep learning models for many disease prediction problems is affected by time-varying covariates, rare incidence, covariate imbalance and delayed diagnosis when using structured electronic health records data. The situation is further exasperated when predicting the risk of one disease on condition of another disease, such as the hepatocellular carcinoma risk among patients with nonalcoholic fatty liver disease due to slow, chronic progression, the scarce of data with both disease conditions and the sex bias of the diseases. The goal of this study is to investigate the extent to which the aforementioned issues influence deep learning performance, and then devised strategies to tackle these challenges. These strategies were applied to improve hepatocellular carcinoma risk prediction among patients with nonalcoholic fatty liver disease.

METHODS

We evaluated two representative deep learning models in the task of predicting the occurrence of hepatocellular carcinoma in a cohort of patients with nonalcoholic fatty liver disease (n = 220,838) from a national EHR database. The disease prediction task was carefully formulated as a classification problem while taking censorship and the length of follow-up into consideration.

RESULTS

We developed a novel backward masking scheme to deal with the issue of delayed diagnosis which is very common in EHR data analysis and evaluate how the length of longitudinal information after the index date affects disease prediction. We observed that modeling time-varying covariates improved the performance of the algorithms and transfer learning mitigated reduced performance caused by the lack of data. In addition, covariate imbalance, such as sex bias in data impaired performance. Deep learning models trained on one sex and evaluated in the other sex showed reduced performance, indicating the importance of assessing covariate imbalance while preparing data for model training.

CONCLUSIONS

The strategies developed in this work can significantly improve the performance of hepatocellular carcinoma risk prediction among patients with nonalcoholic fatty liver disease. Furthermore, our novel strategies can be generalized to apply to other disease risk predictions using structured electronic health records, especially for disease risks on condition of another disease.

Advancing entity recognition in biomedicine via instruction tuning of large language models

MOTIVATION

Large Language Models (LLMs) have the potential to revolutionize the field of Natural Language Processing, excelling not only in text generation and reasoning tasks but also in their ability for zero/few-shot learning, swiftly adapting to new tasks with minimal fine-tuning. LLMs have also demonstrated great promise in biomedical and healthcare applications. However, when it comes to Named Entity Recognition (NER), particularly within the biomedical domain, LLMs fall short of the effectiveness exhibited by fine-tuned domain-specific models. One key reason is that NER is typically conceptualized as a sequence labeling task, whereas LLMs are optimized for text generation and reasoning tasks.

RESULTS

We developed an instruction-based learning paradigm that transforms biomedical NER from a sequence labeling task into a generation task. This paradigm is end-to-end and streamlines the training and evaluation process by automatically repurposing pre-existing biomedical NER datasets. We further developed BioNER-LLaMA using the proposed paradigm with LLaMA-7B as the foundational LLM. We conducted extensive testing on BioNER-LLaMA across three widely recognized biomedical NER datasets, consisting of entities related to diseases, chemicals, and genes. The results revealed that BioNER-LLaMA consistently achieved higher F1-scores ranging from 5% to 30% compared to the few-shot learning capabilities of GPT-4 on datasets with different biomedical entities. We show that a general-domain LLM can match the performance of rigorously fine-tuned PubMedBERT models and PMC-LLaMA, biomedical-specific language model. Our findings underscore the potential of our proposed paradigm in developing general-domain LLMs that can rival SOTA performances in multi-task, multi-domain scenarios in biomedical and health applications.

AVAILABILITY AND IMPLEMENTATION

Datasets and other resources are available at https://github.com/BIDS-Xu-Lab/BioNER-LLaMA.

Study of High-Transverse-Momentum Higgs Boson Production in Association with a Vector Boson in the qqbb Final State with the ATLAS Detector

This Letter presents the first study of Higgs boson production in association with a vector boson (V=W or Z) in the fully hadronic qqbb final state using data recorded by the ATLAS detector at the LHC in proton-proton collisions at sqrt[s]=13  TeV and corresponding to an integrated luminosity of 137  fb^{-1}. The vector bosons and Higgs bosons are each reconstructed as large-radius jets and tagged using jet substructure techniques. Dedicated tagging algorithms exploiting b-tagging properties are used to identify jets consistent with Higgs bosons decaying into bb[over ¯]. Dominant backgrounds from multijet production are determined directly from the data, and a likelihood fit to the jet mass distribution of Higgs boson candidates is used to extract the number of signal events. The VH production cross section is measured inclusively and differentially in several ranges of Higgs boson transverse momentum: 250-450, 450-650, and greater than 650 GeV. The inclusive signal yield relative to the standard model expectation is observed to be μ=1.4_{-0.9}^{+1.0} and the corresponding cross section is 3.1±1.3(stat)_{-1.4}^{+1.8}(syst)  pb.

Spatiotemporal evolution of ultrafast photocarrier dynamics across WS2-ReS2 lateral interface

2D lateral heterostructures possess atomically sharp lateral interfaces, while understanding of their ultrafast photocarrier dynamics from a spatiotemporal viewpoint is rather elusive. In this study, we have investigated the spatiotemporal evolution of photocarrier dynamics across the 1D lateral interface of a WS2-ReS2 2D lateral heterostructure utilizing femtosecond laser pump-probe. The nontrivial band offset across the 1D lateral interface markedly mediates the spatiotemporal photocarrier transfer and transport processes. Subsequently, a hole accumulation region on the WS2 side and an electron accumulation region (1DEG) on the ReS2 side have been spatially identified by correlating ultrafast photocarrier signals. The measured width of the unilateral depletion region is 1360 ± 160 nm. Our work has provided substantial insights into mediated photocarrier dynamics in the 2D lateral heterostructure, which will benefit explorations in 2D interfacial physics and 2D lateral optoelectronic devices.

Ensemble pretrained language models to extract biomedical knowledge from literature

OBJECTIVES

The rapid expansion of biomedical literature necessitates automated techniques to discern relationships between biomedical concepts from extensive free text. Such techniques facilitate the development of detailed knowledge bases and highlight research deficiencies. The LitCoin Natural Language Processing (NLP) challenge, organized by the National Center for Advancing Translational Science, aims to evaluate such potential and provides a manually annotated corpus for methodology development and benchmarking.

MATERIALS AND METHODS

For the named entity recognition (NER) task, we utilized ensemble learning to merge predictions from three domain-specific models, namely BioBERT, PubMedBERT, and BioM-ELECTRA, devised a rule-driven detection method for cell line and taxonomy names and annotated 70 more abstracts as additional corpus. We further finetuned the T0pp model, with 11 billion parameters, to boost the performance on relation extraction and leveraged entites' location information (eg, title, background) to enhance novelty prediction performance in relation extraction (RE).

RESULTS

Our pioneering NLP system designed for this challenge secured first place in Phase I-NER and second place in Phase II-relation extraction and novelty prediction, outpacing over 200 teams. We tested OpenAI ChatGPT 3.5 and ChatGPT 4 in a Zero-Shot setting using the same test set, revealing that our finetuned model considerably surpasses these broad-spectrum large language models.

DISCUSSION AND CONCLUSION

Our outcomes depict a robust NLP system excelling in NER and RE across various biomedical entities, emphasizing that task-specific models remain superior to generic large ones. Such insights are valuable for endeavors like knowledge graph development and hypothesis formulation in biomedical research.

[Risk prediction and function evaluation by T-cell epitope model and expression model of HLA-DPB1 mismatching in unrelated-donor hematopoietic stem cell transplantations]

Objective: To evaluate the risk prediction and assessment function of HLA-DPB1 T-cell epitope (TCE) model and expression model in human leukocyte antigen (HLA)-matched unrelated hematopoietic stem cell transplantation (MUD-HSCT) with HLA-DPB1 mismatching. Methods: A total of 364 (182 pairs) potential MUD-HSCT donors and recipients confirmed by HLA high-resolution typing in Shaanxi Blood Center from 2016 to 2019 were analyzed retrospectively. Of the 182 recipients, there were 121 males and 61 females with an average age of (26.3±14.2) years. Of the 182 donors, there were 148 males and 34 females with an average age of (33.7±7.5) years. Polymerase chain reaction-sequence-based typing (PCR-SBT), next-generation sequencing (NGS) and polymerase chain reaction-sequence specific oligonucleotide probe (PCR-SSO) based on LABScan®3D platform were used for high-resolution typing of HLA-A, B, C, DRB1, DQB1, DPB1 gene, and PCR-SBT was used for single nucleotide polymorphism (SNP) typing. TCE model and expression model were used to predict and evaluate the HLA-DPB1 mismatch pattern and acute graft-versus-host-disease (aGVHD) risk. Results: A total of 26 HLA-DPB1 alleles and their 3'-UTR rs9277534 SNP genotypes were detected in this study population, and two new alleles HLA-DPB1*1052∶01 and HLA-DPB1*1119∶01 were found and officially named. The overall mismatch rate of HLA-DPB1 in MUD-HSCT donors and recipients was 90.66% (165/182). In TCE model, the HLA-DPB1 mismatch rates of permissible mismatch (PM) and non-permissible mismatch (non-PM) were 47.80% (87/182) and 42.86% (78/182), respectively. The non-PM in GvH direction was 13.73% (25/182), and which in HvG direction was 29.12% (53/182). A total of 73 pairs of donors and recipients in TCE model met the evaluation criteria of expression model. Among of TCE PM group, recipient DP5 mismatches accounted for 34.25% (25/73) were predicted as aGVHD high risk according to expression model. For the TCE non-PM group, both the recipient DP2 mismatches of 6.85% (5/73) and recipient DP5 mismatches of 10.86% (8/73) were predicted to be at high risk for aGVHD. Risk prediction by TCE model and expression model was 27.27% concordant and 16.97% unconcordant. Conclusions: TCE model and expression model are effective tools to predict aGVHD risk of MUD-HSCT. Comprehensive application of the two models is helpful to the hierarchical assessment of HSCT risk.

Integrated single-cell sequencing, spatial transcriptome sequencing and bulk RNA sequencing highlights the molecular characteristics of parthanatos in gastric cancer

BACKGROUND

Parthanatos is a novel programmatic form of cell death based on DNA damage and PARP-1 dependency. Nevertheless, its specific role in the context of gastric cancer (GC) remains uncertain.

METHODS

In this study, we integrated multi-omics algorithms to investigate the molecular characteristics of parthanatos in GC. A series of bioinformatics algorithms were utilized to explore clinical heterogeneity of GC and further predict the clinical outcomes.

RESULTS

Firstly, we conducted a comprehensive analysis of the omics features of parthanatos in various human tumors, including genomic mutations, transcriptome expression, and prognostic relevance. We successfully identified 7 cell types within the GC microenvironment: myeloid cell, epithelial cell, T cell, stromal cell, proliferative cell, B cell, and NK cell. When compared to adjacent non-tumor tissues, single-cell sequencing results from GC tissues revealed elevated scores for the parthanatos pathway across multiple cell types. Spatial transcriptomics, for the first time, unveiled the spatial distribution characteristics of parthanatos signaling. GC patients with different parthanatos signals often exhibited distinct immune microenvironment and metabolic reprogramming features, leading to different clinical outcomes. The integration of parthanatos signaling and clinical indicators enabled the creation of novel survival curves that accurately assess patients' survival times and statuses.

CONCLUSIONS

In this study, the molecular characteristics of parthanatos' unicellular and spatial transcriptomics in GC were revealed for the first time. Our model based on parthanatos signals can be used to distinguish individual heterogeneity and predict clinical outcomes in patients with GC.

COVID-19 vaccination and the risk of autoimmune diseases: a Mendelian randomization study

Background

In recent times, reports have emerged suggesting that a variety of autoimmune disorders may arise after the coronavirus disease 2019 (COVID-19) vaccination. However, causality and underlying mechanisms remain unclear.

Methods

We collected summary statistics of COVID-19 vaccination and 31 autoimmune diseases from genome-wide association studies (GWAS) as exposure and outcome, respectively. Random-effects inverse variance weighting (IVW), MR Egger, weighted median, simple mode, and weighted mode were used as analytical methods through Mendelian randomization (MR), and heterogeneity and sensitivity analysis were performed.

Results

We selected 72 instrumental variables for exposure (p < 5 × 10-6; r2 < 0.001, genetic distance = 10,000 kb), and MR analyses showed that COVID-19 vaccination was causally associated with an increased risk of multiple sclerosis (MS) (IVW, OR: 1.53, 95% CI: 1.065-2.197, p = 0.026) and ulcerative colitis (UC) (IVW, OR: 1.00, 95% CI: 1.000-1.003, p = 0.039). If exposure was refined (p < 5 × 10-8; r2 < 0.001, genetic distance = 10,000 kb), the associations became negative. No causality was found for the remaining outcomes. These results were robust to sensitivity and heterogeneity analyses.

Conclusion

Our study provided potential evidence for the impact of COVID-19 vaccination on the risk of MS and UC occurrence, but it lacks sufficient robustness, which could provide a new idea for public health policy.

Measurement of the Centrality Dependence of the Dijet Yield in p+Pb Collisions at sqrt[s_{NN}]=8.16 TeV with the ATLAS Detector

ATLAS measured the centrality dependence of the dijet yield using 165  nb^{-1} of p+Pb data collected at sqrt[s_{NN}]=8.16  TeV in 2016. The event centrality, which reflects the p+Pb impact parameter, is characterized by the total transverse energy registered in the Pb-going side of the forward calorimeter. The central-to-peripheral ratio of the scaled dijet yields, R_{CP}, is evaluated, and the results are presented as a function of variables that reflect the kinematics of the initial hard parton scattering process. The R_{CP} shows a scaling with the Bjorken x of the parton originating from the proton, x_{p}, while no such trend is observed as a function of x_{Pb}. This analysis provides unique input to understanding the role of small proton spatial configurations in p+Pb collisions by covering parton momentum fractions from the valence region down to x_{p}∼10^{-3} and x_{Pb}∼4×10^{-4}.

Expression of apelin‑13 and its negative correlation with TGF‑β1 in patients with diabetic kidney disease

Diabetic kidney disease (DKD) is a severe microvascular complication of diabetes, one key feature of which includes renal fibrosis. As apelin is an adipokine closely related to diabetes, the present study aimed to evaluate apelin-13 expression levels and the relationship between apelin-13 and disease indicators in patients with diabetic kidney disease (DKD). The present case-control study enrolled 70 patients with diabetes, including 31 with diabetic kidney disease (DKD group), 39 without DKD (non-DKD group) and 30 healthy controls. The levels of serum apelin-13 and TGF-β1, the key driver of renal fibrosis, were determined by ELISA. Additionally, age, mean disease duration, weight, blood pressure, fasting blood glucose, triglyceride, low-density lipoprotein cholesterol, high-density lipoprotein, cholesterol, urea nitrogen, blood creatinine and 24-hour urinary total protein (24-h UTP) were recorded. The results demonstrated that apelin-13 and TGF-β1 expression levels, age, blood pressure, fasting blood glucose, cholesterol and blood urea nitrogen levels were significantly higher in patients with diabetes compared with the healthy controls (P<0.05). Moreover, apelin-13 and TGF-β1 expression levels, mean disease duration, systolic pressure, blood creatinine, blood urea nitrogen and 24-h UTP were significantly higher in the DKD group compared with the non-DKD group (P<0.05). The estimated glomerular filtration rate (eGFR) was significantly reduced in the DKD group compared with the non-DKD group (P<0.05). Correlation analysis demonstrated a negative correlation between apelin-13 and eGFR expression and a positive correlation between apelin-13 expression and 24-h UTP in both the DKD and non-DKD groups (P<0.05). A negative correlation was also demonstrated between apelin-13 and TGF-β1 expression levels in the DKD group and non-DKD groups (both P<0.05). In conclusion, apelin-13 and TGF-β1 expression levels were significantly higher in the DKD group compared with those in the non-DKD group. Additionally, apelin-13 expression was negatively correlated with TGF-β1 expression in the DKD and non-DKD groups. Therefore, apelin-13 could potentially be used in the future as an indicator of renal fibrosis or destruction in patients with DKD. The present trial was retrospectively registered in the Chinese Clinical Trial Registry (trial registration no. ChiCTR2200060945) on 14.06.2022.

A scoping review of fair machine learning techniques when using real-world data

OBJECTIVE

The integration of artificial intelligence (AI) and machine learning (ML) in health care to aid clinical decisions is widespread. However, as AI and ML take important roles in health care, there are concerns about AI and ML associated fairness and bias. That is, an AI tool may have a disparate impact, with its benefits and drawbacks unevenly distributed across societal strata and subpopulations, potentially exacerbating existing health inequities. Thus, the objectives of this scoping review were to summarize existing literature and identify gaps in the topic of tackling algorithmic bias and optimizing fairness in AI/ML models using real-world data (RWD) in health care domains.

METHODS

We conducted a thorough review of techniques for assessing and optimizing AI/ML model fairness in health care when using RWD in health care domains. The focus lies on appraising different quantification metrics for accessing fairness, publicly accessible datasets for ML fairness research, and bias mitigation approaches.

RESULTS

We identified 11 papers that are focused on optimizing model fairness in health care applications. The current research on mitigating bias issues in RWD is limited, both in terms of disease variety and health care applications, as well as the accessibility of public datasets for ML fairness research. Existing studies often indicate positive outcomes when using pre-processing techniques to address algorithmic bias. There remain unresolved questions within the field that require further research, which includes pinpointing the root causes of bias in ML models, broadening fairness research in AI/ML with the use of RWD and exploring its implications in healthcare settings, and evaluating and addressing bias in multi-modal data.

CONCLUSION

This paper provides useful reference material and insights to researchers regarding AI/ML fairness in real-world health care data and reveals the gaps in the field. Fair AI/ML in health care is a burgeoning field that requires a heightened research focus to cover diverse applications and different types of RWD.

Staged Immediate Nipple Reconstruction With Tube Flap in Immediate Deep Inferior Epigastric Artery Perforator Flap Breast Reconstruction

BACKGROUND

In the setting of immediate breast reconstruction by deep inferior epigastric artery perforator (DIEP) flap, the excessive DIEP flap skin is de-epithelialized and then buried under the mastectomy skin. In this study, by virtue of tube flap technique, we hypothesize that the skin supposed to be abandoned could be transferred to the apex of reconstructed breast mound for nipple reconstruction.

METHODS

A total of 60 female patients were recruited between January 2019 and December 2020. All these patients underwent mastectomy including nipple-areola complex and immediate DIEP flap breast reconstruction. A ladder-shaped pedicled flap was raised from the DIEP flap and rolled into a tube. The free end of tube flap was inset into the future nipple position of the reconstructed breast mound 1 week later. After revascularization for 1 month, we divided the previous pedicle and used the tube on the apex of the breast mound to recreate a new nipple.

RESULTS

All reconstructed breasts and nipples survived well postoperatively. The average nipple projection was 12.5 ± 2.0 mm immediately after the surgery, which gradually decreased to 9.4 ± 1.5 mm at 1-year follow-up, with the projection loss from the initial measurement as 24.9% ± 1.8%. In total, 51 patients considered the overall impression of breast and nipple reconstruction to be very good or good.

CONCLUSIONS

We provided an ideal technique that could improve the maintenance of reconstructed nipple projection and have aesthetically acceptable outcomes, without DIEP flap tissue loss, breast mound distortion, or additional scars.

Divinyl sulfone, an oxidative metabolite of sulfur mustard, induces caspase-independent pyroptosis in hepatocytes

Sulfur mustard (SM) is a highly toxic blister agent which has been used many times in several wars and conflicts and caused heavy casualties. Ease of production and lack of effective therapies make SM a potential threat to public health. SM intoxication causes severe damage on various target organs, such as the skin, eyes, and lungs. In addition, SM exposure can also lead to hepatotoxicity and severe liver injuries. However, despite decades of research, the molecular mechanism underlying SM-induced liver damage remains obscure. SM can be converted into various products via complex hepatic metabolism in vivo. There are some pieces of evidence that one of the oxidation products of SM, divinyl sulfone (DVS), exhibits even more significant toxicity than SM. Nevertheless, the molecular toxicology of DVS is still hardly known. In the present study, we confirmed that DVS is even more toxic than SM in the human hepatocellular carcinoma cell line HepG2. Further mechanistic study revealed that DVS exposure (200 μM) promotes pyroptosis in HepG2 cells, while SM (400 μM) mainly induces apoptosis. DVS induces gasdermin D (GSDMD) mediated pyroptosis, which is independent of caspases activation but depends on the large amounts of reactive oxygen species (ROS) and severe oxidative stress produced during DVS exposure. Our findings may provide novel insights for understanding the mechanism of SM poisoning and may be helpful to discover promising therapeutic strategies for SM intoxication.

Clinical evaluation of resorbable polylactic acid (PLA) intracanal posts for primary incisor restoration. Randomized controlled clinical trial

This randomized, controlled clinical trial compares the clinical performance of glass-fibre and resorbable polylactic acid (PLA) intracanal posts used to restore carious primary incisors in young patients. The study sample includes 180 primary upper central incisors of 90 children aged 3 to 4 years. All patients were randomly divided into two equal groups of 45 children who received PLA and glass-fibre (GFP) intracanal posts. The clinical assessment of incisor restorations was carried out immediately upon completion and at months 3, 6 and 12 according to the following criteria: anatomical form, marginal adaptation, surface roughness, marginal pigmentation, colour match, secondary caries and contact point. The Gingival Index (GI), the Bleeding Index (Cowell modification; mBI), and bite force (BF) were measured. At the 3-month follow-up, the occlusal BF of patients who received PLA posts was higher than the baseline; the GI and mBI scores were lower, by contrast (p < 0.05). This tendency was even more pronounced 6 and 12 months after the restoration. The incidence of side effects or symptoms (apical inflammation, cervical fracture, loosening of the crown) after the PLA posts was significantly lower than after the GFP (p < 0.05). No statistically significant differences were present between the two groups with respect to colour matching, anatomical form, marginal adaptation, marginal pigmentation, surface roughness, occlusal contact and secondary caries. Based on the results, applying PLA intracanal posts and cyanoacrylate to residual anterior crowns in young children can improve their gingival health, reduce side effects, and increase the likelihood of successful restoration.

Angle-Resolved Optical Imaging of Interlayer Rotations in Twisted Bilayer Graphene

Twisted bilayer graphene (TBG) is a prototypical layered material whose properties are strongly correlated to interlayer coupling. The two stacked graphene layers with distinct orientations are investigated to generate peculiar optical and electronic phenomena. Thus, the rapid, reliable, and nondestructive twist angle identification technique is of essential importance. Here, we integrated the white light reflection spectra (WLRS), the Raman spectroscopy, and the transmission electron microscope (TEM) to propose a facile RGB optical imaging technique that identified the twist angle of the TBG in a large area intuitively with high efficiency. The RGB technique established a robust correlation between the interlayer rotation angle and the contrast difference in the RGB color channels of a standard optical image. The angle-resolved optical behavior is attributed to the absorption resonance matching with the separation of van Hove singularities in the density of states of the TBG. Our study thus developed a route to identify the rotation angle of stacked bilayer graphene by means of a straightforward optical method, which can be further applied in other stacked van der Waals layered materials.

ALOX5 drives the pyroptosis of CD4+ T cells and tissue inflammation in rheumatoid arthritis

Pyroptosis, an inflammatory form of programmed cell death, is linked to the pathology of rheumatoid arthritis (RA). Here, we investigated the molecular mechanism underlying pyroptosis in T cells isolated from patients with RA. Compared with healthy individuals, patients with RA had more pyroptotic CD4+ T cells in blood and synovia, which correlated with clinical measures of disease activity. Moreover, the mRNA expression and protein abundance of arachidonate 5-lipoxygenase (ALOX5), which converts arachidonic acid to leukotriene A4 (LTA4), were increased in CD4+ T cells from patients with RA and, among patients with RA, were lowest in those in clinical remission. Knockdown or pharmacological inhibition of ALOX5 suppressed CD4+ T cell pyroptosis and improved symptoms in two rodent models of RA. Mechanistically, the increase in ALOX5 activity in RA CD4+ T cells enhanced the production of the LTA4 derivative LTB4, which stimulated Ca2+ influx through ORAI3 channels, leading to the activation of NLRP3 inflammasomes and pyroptosis. Our findings reveal a role for ALOX5 in RA and provide a molecular basis for further exploring the clinical utility of ALOX5 inhibition in RA and for using ALOX5 as a biomarker to distinguish active disease and remission in RA.