OSBPL2 compound heterozygous variants cause dyschromatosis, ichthyosis, deafness and atopic disease syndrome.

PURPOSE: In this study, we identified and diagnosed a novel inherited condition called Dyschromatosis, Ichthyosis, Deafness, and Atopic Disease (DIDA) syndrome. We present a series of studies to clarify the pathogenic variants and specific mechanism. METHODS: Exome sequencing and Sanger sequencing was conducted in affected and unaffected family members. A variety of human and cell studies were performed to explore the pathogenic process of keratosis. RESULTS: Our finding indicated that DIDA syndrome was caused by compound heterozygous variants in the oxysterol-binding protein-related protein 2 (OSBPL2) gene. Furthermore, our findings revealed a direct interaction between OSBPL2 and Phosphoinositide phospholipase C-beta-3 (PLCB3), a key player in hyperkeratosis. OSBPL2 effectively inhibits the ubiquitylation of PLCB3, thereby stabilizing PLCB3. Conversely, OSBPL2 variants lead to enhanced ubiquitination and subsequent degradation of PLCB3, leading to epidermal hyperkeratosis, characterized by aberrant proliferation and delayed terminal differentiation of keratinocytes. CONCLUSIONS: Our study not only unveiled the association between OSBPL2 variants and the newly identified DIDA syndrome but also shed light on the underlying mechanism.

Acceptance model of new energy vehicles based on PLS-SEM model.

The current energy crisis is worsening worldwide, and in China, urban expansion and per capita vehicle ownership have created a growing energy imbalance and increased pressure to reduce carbon emissions.The popularization of new energy vehicles (NEVs) can provide a step forward to solving energy shortage problems, environmental pollution, and global warming. In 2022, the average penetration rate, which is ratio of new energy vehicle sales to vehicle sales, is just 19.1 %. This paper analysed the reasons for the differences in the penetration rates of new energy vehicles in China's 269 prefecture-level cities, using a Geo Detector approach, and the results showed that the level of economic development, the average annual temperature difference, the density of charging piles, the charging price and the number of population all had significant effects(q＞0.12) on the penetration rate. Based on the above studies, a questionnaire was used to investigate the public's acceptance of new energy vehicles in Xinjiang Uygur Autonomous Region, and a PLS-SEM regression analysis was conducted. The results showed that men, young people and people with a certain level of basic education were 5 % more likely to accept new energy vehicles.Unlike previous studies, perceived cost had no significant correlation with the acceptance of new energy vehicles. Perceived risk had a significant negative correlation with the acceptance of new energy vehicles,the path coefficient is -0.1.The acceptance of new energy vehicles was significantly and positively correlated with vehicle quality and service, the public's understanding of new energy vehicles, and subjective norms, their average path coefficients are above 0.1. We argues that the government should maintain a certain level of promotion of new energy vehicles and accelerate the construction of charging piles, based on the aforementioned results.

Speckle Variance Photoacoustic Microscopy for Microhemodynamic Imaging.

Relying on the strong optical absorption of hemoglobin to pulsed laser energy, photoacoustic microscopy provides morphological and functional information on microvasculature label-freely. Here, we propose speckle variance photoacoustic microscopy (SV-PAM), which harnesses intrinsic imaging contrast from temporal-varied photoacoustic signals of moving red blood cells in blood vessels, for recovering three-dimension hemodynamic images down to capillary-level resolution within the microcirculatory tissue beds in vivo. Calculating the speckle variance of consecutive photoacoustic B-scan frames acquired at the same lateral position enables accurate identification of blood perfusion and occlusion, which provides interpretations of dynamic blood flow in the microvasculature, in addition to the microvascular anatomic structures. We demonstrate high-resolution hemodynamic imaging of vascular occlusion and reperfusion in the microvasculature of mice ears in vivo. The results suggest that our SV-PAM is potentially invaluable for biomedical hemodynamic investigations, for example, imaging ischemic stroke and hemorrhagic stroke.

Lusutrombopag as a Repurposing Drug in Combination with Aminoglycosides against Vancomycin-Resistant Enterococcus.

Due to the widespread abuse of antibiotics, drug resistance in Enterococcus has been increasing. However, the speed of antibiotic discovery cannot keep pace with the acquisition of bacterial resistance. Thus, drug repurposing is a proposed strategy to solve the crises. Lusutrombopag (LP) has been approved as a thrombopoietin receptor agonist by the Food and Drug Administration. This study demonstrated that LP exhibited significant antimicrobial activities against vancomycin-resistant Enterococcus in vitro with rare resistance occurrence. Further, LP combined with tobramycin exhibited synergistic antimicrobial effects in vitro and in vivo against Enterococcus. No in vitro or in vivo detectable toxicity was observed when using LP. Mechanism studies indicated that the disrupted proton motive force may account for LP's antimicrobial activity. In summary, these results demonstrate that LP has the previously undocumented potential to serve as an antibacterial agent against refractory infections caused by Enterococcus.

Tandem Vinylogous Aldol and Intramolecular [2 + 2] Cycloaddition toward Benzocyclobutenes by UV Light Photocatalysis.

A facile and efficient radical tandem vinylogous aldol and intramolecular [2 + 2] cycloaddition reaction for direct synthesis of cyclobutane-containing benzocyclobutenes (BCBs) under extremely mild conditions without using any photocatalysts is reported. This approach exhibited definite compatibility with functional groups and afforded new BCBs with excellent regioselectivity and high yields. Moreover, detailed mechanism studies were carried out both experimentally and theoretically. The readily accessible, low-cost, and ecofriendly nature of the developed strategy will endow it with attractive applications in organic and medicinal chemistry.

Human umbilical cord mesenchymal stem cell-derived exosomes ameliorate renal fibrosis in diabetic nephropathy by targeting Hedgehog/SMO signaling.

Diabetic nephropathy (DN) is the leading cause of end-stage renal disease globally. Currently, there are no effective drugs for the treatment of DN. Although several studies have reported the therapeutic potential of mesenchymal stem cells, the underlying mechanisms remain largely unknown. Here, we report that both human umbilical cord MSCs (UC-MSCs) and UC-MSC-derived exosomes (UC-MSC-exo) attenuate kidney damage, and inhibit epithelial-mesenchymal transition (EMT) and renal fibrosis in streptozotocin-induced DN rats. Strikingly, the Hedgehog receptor, smoothened (SMO), was significantly upregulated in the kidney tissues of DN patients and rats, and positively correlated with EMT and renal fibrosis. UC-MSC and UC-MSC-exo treatment resulted in decrease of SMO expression. In vitro co-culture experiments revealed that UC-MSC-exo reduced EMT of tubular epithelial cells through inhibiting Hedgehog/SMO pathway. Collectively, UC-MSCs inhibit EMT and renal fibrosis by delivering exosomes and targeting Hedgehog/SMO signaling, suggesting that UC-MSCs and their exosomes are novel anti-fibrotic therapeutics for treating DN.

Mild Therapeutic Hypothermia Alleviated Myocardial Ischemia/Reperfusion Injury via Targeting SLC25A10 to Suppress Mitochondrial Apoptosis.

Myocardial ischemia/reperfusion injury (MI/RI) is identified as a severe vascular emergency, and the treatment strategy of MI/RI still needs further improvement. The present study aimed to investigate the potential effects of mild therapeutic hypothermia (MTH) on MI/RI and underlying mechanisms. In ischemia/reperfusion (I/R) rats, MTH treatment significantly improved myocardial injury, attenuated myocardial infarction, and inhibited the mitochondrial apoptosis pathway. The results of proteomics identified SLC25A10 as the main target of MTH treatment. Consistently, SLC25A10 expressions in I/R rat myocardium and hypoxia and reoxygenation (H/R) cardiomyocytes were significantly suppressed, which was effectively reversed by MTH treatment. In H/R cardiomyocytes, MTH treatment significantly improved cell injury, mitochondrial dysfunction, and inhibited the mitochondrial apoptosis pathway, which were partially reversed by SLC25A10 deletion. These findings suggested that MTH treatment could protect against MI/RI by modulating SLC25A10 expression to suppress mitochondrial apoptosis pathway, providing new theoretical basis for clinical application of MTH treatment for MI/RI.

A long-acting FGF21 attenuates metabolic dysfunction-associated steatohepatitis-related fibrosis by modulating NR4A1-mediated Ly6C phenotypic switch in macrophages.

BACKGROUND AND PURPOSE: Because of the absence of effective therapies for metabolic dysfunction-associated steatohepatitis (MASH), there is a rising interest in fibroblast growth factor 21 (FGF21) analogues due to their potential anti-fibrotic activities in MASH treatment. PsTag-FGF21, a long-acting FGF21 analogue, has demonstrated promising therapeutic effects in several MASH mouse models. However, its efficacy and mechanism against MASH-related fibrosis remain less well defined, compared with the specific mechanisms through which FGF21 improves glucose and lipid metabolism. EXPERIMENTAL APPROACH: The effectiveness of PsTag-FGF21 was evaluated in two MASH-fibrosis models. Co-culture systems involving macrophages and hepatic stellate cells (HSCs) were employed for further assessment. Hepatic macrophages were selectively depleted by administering liposome-encapsulated clodronate via tail vein injections. RNA sequencing and cytokine profiling were conducted to identify key factors involved in macrophage-HSC crosstalk. KEY RESULTS: We first demonstrated the significant attenuation of hepatic fibrosis by PsTag-FGF21 in two MASH-fibrosis models. Furthermore, we highlighted the crucial role of macrophage phenotypic switch in PsTag-FGF21-induced HSC deactivation. FGF21 was demonstrated to regulate macrophages in a PsTag-FGF21-like manner. NR4A1, a nuclear factor which is notably down-regulated in human livers with MASH, was identified as a mediator responsible for PsTag-FGF21-induced phenotypic switch. Transcriptional control over insulin-like growth factor 1, a crucial factor in macrophage-HSC crosstalk, was exerted by the intrinsically disordered region domain of NR4A1. CONCLUSION AND IMPLICATIONS: Our results have elucidated the previously unclear mechanisms through which PsTag-FGF21 treats MASH-related fibrosis and identified NR4A1 as a potential therapeutic target for fibrosis.

A survey of recent methods for addressing AI fairness and bias in biomedicine.

OBJECTIVES: Artificial intelligence (AI) systems have the potential to revolutionize clinical practices, including improving diagnostic accuracy and surgical decision-making, while also reducing costs and manpower. However, it is important to recognize that these systems may perpetuate social inequities or demonstrate biases, such as those based on race or gender. Such biases can occur before, during, or after the development of AI models, making it critical to understand and address potential biases to enable the accurate and reliable application of AI models in clinical settings. To mitigate bias concerns during model development, we surveyed recent publications on different debiasing methods in the fields of biomedical natural language processing (NLP) or computer vision (CV). Then we discussed the methods, such as data perturbation and adversarial learning, that have been applied in the biomedical domain to address bias. METHODS: We performed our literature search on PubMed, ACM digital library, and IEEE Xplore of relevant articles published between January 2018 and December 2023 using multiple combinations of keywords. We then filtered the result of 10,041 articles automatically with loose constraints, and manually inspected the abstracts of the remaining 890 articles to identify the 55 articles included in this review. Additional articles in the references are also included in this review. We discuss each method and compare its strengths and weaknesses. Finally, we review other potential methods from the general domain that could be applied to biomedicine to address bias and improve fairness. RESULTS: The bias of AIs in biomedicine can originate from multiple sources such as insufficient data, sampling bias and the use of health-irrelevant features or race-adjusted algorithms. Existing debiasing methods that focus on algorithms can be categorized into distributional or algorithmic. Distributional methods include data augmentation, data perturbation, data reweighting methods, and federated learning. Algorithmic approaches include unsupervised representation learning, adversarial learning, disentangled representation learning, loss-based methods and causality-based methods.

Hippocampal replay sequence governed by spontaneous brain-wide dynamics.

Neurons in the hippocampus exhibit spontaneous spiking activity during rest that appears to recapitulate previously experienced events. While this replay activity is frequently linked to memory consolidation and learning, the underlying mechanisms are not well understood. Recent large-scale neural recordings in mice have demonstrated that resting-state spontaneous activity is expressed as quasi-periodic cascades of spiking activity that pervade the forebrain, with each cascade engaging a high proportion of recorded neurons. Hippocampal ripples are known to be coordinated with cortical dynamics; however, less is known about the occurrence of replay activity relative to other brain-wide spontaneous events. Here we analyzed responses across the mouse brain to multiple viewings of natural movies, as well as subsequent patterns of neural activity during rest. We found that hippocampal neurons showed time-selectivity, with individual neurons responding consistently during particular moments of the movie. During rest, the population of time-selective hippocampal neurons showed both forward and time-reversed replay activity that matched the sequence observed in the movie. Importantly, these replay events were strongly time-locked to brain-wide spiking cascades, with forward and time-reversed replay activity associated with distinct cascade types. Thus, intrinsic hippocampal replay activity is temporally structured according to large-scale spontaneous physiology affecting areas throughout the forebrain. These findings shed light on the coordination between hippocampal and cortical circuits thought to be critical for memory consolidation.

Prognostic Value of a Combined Nomogram Model Integrating 3-Dimensional Deep Learning and Radiomics for Head and Neck Cancer.

OBJECTIVE: The preoperative prediction of the overall survival (OS) status of patients with head and neck cancer (HNC) is significant value for their individualized treatment and prognosis. This study aims to evaluate the impact of adding 3D deep learning features to radiomics models for predicting 5-year OS status. METHODS: Two hundred twenty cases from The Cancer Imaging Archive public dataset were included in this study; 2212 radiomics features and 304 deep features were extracted from each case. The features were selected by univariate analysis and the least absolute shrinkage and selection operator, and then grouped into a radiomics model containing Positron Emission Tomography /Computed Tomography (PET/CT) radiomics features score, a deep model containing deep features score, and a combined model containing PET/CT radiomics features score +3D deep features score. TumorStage model was also constructed using initial patient tumor node metastasis stage to compare the performance of the combined model. A nomogram was constructed to analyze the influence of deep features on the performance of the model. The 10-fold cross-validation of the average area under the receiver operating characteristic curve and calibration curve were used to evaluate performance, and Shapley Additive exPlanations (SHAP) was developed for interpretation. RESULTS: The TumorStage model, radiomics model, deep model, and the combined model achieved areas under the receiver operating characteristic curve of 0.604, 0.851, 0.840, and 0.895 on the train set and 0.571, 0.849, 0.832, and 0.900 on the test set. The combined model showed better performance of predicting the 5-year OS status of HNC patients than the radiomics model and deep model. The combined model was shown to provide a favorable fit in calibration curves and be clinically useful in decision curve analysis. SHAP summary plot and SHAP The SHAP summary plot and SHAP force plot visually interpreted the influence of deep features and radiomics features on the model results. CONCLUSIONS: In predicting 5-year OS status in patients with HNC, 3D deep features could provide richer features for combined model, which showed outperformance compared with the radiomics model and deep model.

Plasmonic Gold Nanostar-Based Probes with Distance-Dependent Plasmon-Enhanced Fluorescence for Ultrasensitive DNA Methyltransferase Assay.

The ultrasensitive DNA methyltransferase (Dam MTase) assay is of high significance for biomedical research and clinical diagnosis because of its profound effect on gene regulation. However, detection sensitivity is still limited by shortcomings, including photobleaching and weak signal intensities of conventional fluorophores at low concentrations. Plasmonic nanostructures with ultrastrong electromagnetic fields and fluorescence enhancement capability that can overcome these intrinsic defects hold great potential for ultrasensitive bioanalysis. Herein, a silica-coated gold nanostars (Au NSTs@SiO2)-based plasmon-enhanced fluorescence (PEF) probe with 20 "hot spots" was developed for ultrasensitive detection of Dam MTase. Here, the Dam Mtase assay was achieved by detecting the byproduct PPi of the rolling circle amplification reaction. It is worth noting that, benefiting from the excellent fluorescence enhancement capability of Au NSTs originating from their 20 "hot spots", the detection limit of Dam Mtase was reduced by nearly 105 times. Moreover, the proposed Au NST-based PEF probe enabled versatile evaluation of Dam MTase inhibitors as well as endogenous Dam MTase detection in GW5100 and JM110 Escherichia coli cell lysates, demonstrating its potential in biomedical analysis.

Fabrication of hyaluronic acid-altered gold complex delivery for head and neck squamous cell carcinoma therapy with high antitumor efficacy and low in vivo toxicity.

The use of multifunctional nanomedicines in the treatment of tumors is gaining popularity. Here, we constructed a nanodrug delivery system (HA/Au-PDA@CZT) that targets tumors and responds to pH and near-infrared (NIR) dual stimuli. By precisely interacting with an overexpressed CD44 receptor in specific cancer cells, hyaluronic acid (HA) is coated on the Au-PDA NP surface for tumor-targeting abilities. When exposed to NIR radiation, polydopamine (PDA) and gold nanoshells exhibit exceptional photothermal performance that has the potential to both accelerate and kill HLAC 78 head and neck squamous cell carcinoma cells. Antitumor investigations conducted in vivo and in vitro demonstrated that nanomedicine had remarkable synergistic benefits with chemotherapy and photothermal treatment. Only 25.2% of the cells in the HA/Au-PDA@CZT with a NIR irradiation group were viable. Any group's lowest tumor volume was shown in the tumor mice subjected to HA/Au-PDA@CZT with NIR at 0.3 ± 0.1. Consequently, for synergistic chemo-photothermal therapy, our logically designed nanoplatform would be the potential for a head and neck squamous tumor-targeting drug delivery system.

Utilizing rubidium chloride as an effective and stable interface modification layer for high-efficiency solar cells.

The presence of interface defects between the perovskite layer and the underlying substrate has a significant impact on the power conversion efficiency (PCE) and stability of perovskite solar cells (PSCs). S n O 2 thin films are employed in PSCs as electron transport layers to achieve high PCE. However, the significant lattice mismatch between S n O 2 and the perovskite material leads to a large number of uncoordinated defects at the interface between perovskite and substrate, resulting in recombination losses at the interface. In this study, rubidium chloride (RbCl) was introduced as the interface modification layer between the perovskite layer and the S n O 2 electron transport layer to enhance the PCE of PSCs. The research showed that the RbCl interface modification layer effectively passivated the under-coordinated defects of Sn ions and optimized the energy level alignment between the perovskite layer and the S n O 2 film. Moreover, the fabricated PSCs exhibited an open-circuit voltage of 1.11 V and a power conversion efficiency of 21.64%. Furthermore, the device maintained 80% of initial efficiency after storage for 30 days in an inert gas environment and 60% of the value after storage for 30 days in ambient air.

The Gothenburg Trismus Questionnaire in China: Cross-cultural adaptation and measurement invariance.

OBJECTIVES: The Gothenburg Trismus Questionnaire (GTQ) is a comprehensive scale for screening and assessing trismus in head and neck (H&N) cancer and temporomandibular joint disorders (TMD) patients. This study aimed to translate and cross-culturally adapt the GTQ in China, and to test its measurement invariance. METHODS: This study comprised 278 H&N cancer, 245 TMD, and 507 control patients. Internal consistency and test-retest reliability were tested to assess the GTQ's reliability. The validity was evaluated through composite reliability (CR), average variance extracted (AVE), and correlation tests. Multi-group confirmatory factor analysis (CFA) was used to investigate the GTQ's measurement invariance across clinical status and gender. T tests were employed to compare score differences across clinical status and gender. RESULTS: The Chinese version of GTQ scale shows excellent internal consistency and test-retest reliability. The CR, AVE, and correlation values demonstrate the good validity of GTQ. The multi-group CFA supported configural invariance across clinical status but not metric invariance, while it supported strict invariance across gender. Additionally, t tests revealed that patients with H&N cancer and TMD scored higher than the control group, while males scored higher than females. CONCLUSIONS: The Chinese version of GTQ serves as an effective tool for screening and assessing trismus.

Progress in Biomarkers Related to Biliary Atresia.

Biliary atresia (BA) is a congenital cholestatic disease that can seriously damage children's liver function. It is one of the main reasons for liver transplantation in children. Early diagnosis of BA is crucial to the prognosis of patients, but there is still a lack of reliable non-invasive diagnostic methods. Additionally, as some children are in urgent need of liver transplantation, evaluating the stage of liver fibrosis and postoperative native liver survival in children with BA using a straightforward, efficient, and less traumatic method is a major focus of doctors. In recent years, an increasing number of BA-related biomarkers have been identified and have shown great potential in the following three aspects of clinical practice: diagnosis, evaluation of the stage of liver fibrosis, and prediction of native liver survival. This review focuses on the pathophysiological function and clinical application of three novel BA-related biomarkers, namely MMP-7, FGF-19, and M2BPGi. Furthermore, progress in well-known biomarkers of BA such as gamma-glutamyltransferase, circulating cytokines, and other potential biomarkers is discussed, aiming to provide a reference for clinical practice.

Quantitative Comparison of Liver Volume, Proton Density Fat Fraction, and Time Burden between Automatic Whole Liver Segmentation and Manual Sampling MRI Strategies for Diagnosing Metabolic Dysfunction-associated Steatotic Liver Disease in Obese Patients.

BACKGROUND: The performance of automatic liver segmentation and manual sampling MRI strategies needs be compared to determine interchangeability. OBJECTIVE: To compare automatic liver segmentation and manual sampling strategies (manual whole liver segmentation and standardized manual region of interest) for performance in quantifying liver volume and MRI-proton density fat fraction (MRI-PDFF), identifying steatosis grade, and time burden. METHODS: Fifty patients with obesity who underwent liver biopsy and MRI between December 2017 and November 2018 were included. Sampling strategies included automatic and manual whole liver segmentation and 4 and 9 large regions of interest. Intraclass correlation coefficient (ICC), Bland-Altman, linear regression, receiver operating characteristic curve, and Pearson correlation analyses were performed. RESULTS: Automatic whole liver segmentation liver volume and manual whole liver segmentation liver volume showed excellent agreement (ICC=0.97), high correlation (R2=0.96), and low bias (3.7%, 95% limits of agreement, -4.8%, 12.2%) in liver volume. There was the best agreement (ICC=0.99), highest correlation (R2=1.00), and minimum bias (0.84%, 95% limits of agreement, -0.20%, 1.89%) between automated whole liver segmentation MRI-PDFF and manual whole liver segmentation MRI-PDFF. There was no difference of each paired comparison of receiver operating characteristic curves for detecting steatosis (P=0.07-1.00). The minimum time burden for automatic whole liver segmentation was 0.32 s (0.32-0.33 s). CONCLUSION: Automatic measurement has similar effects to manual measurement in quantifying liver volume, MRI-PDFF, and detecting steatosis. Time burden of automatic whole liver segmentation is minimal among all sampling strategies. Manual measurement can be replaced by automatic measurement to improve quantitative efficiency.

Sex-Specific Effect of MTSS1 Downregulation on Dilated Cardiomyopathy.

MTSS1 (metastasis suppressor 1) is an I-BAR protein that regulates cytoskeleton dynamics through interactions with actin, Rac, and actin-associated proteins. In a prior study, we identified genetic variants in a cardiac-specific enhancer upstream of MTSS1 that reduce human left ventricular (LV) MTSS1 expression and associate with protection against dilated cardiomyopathy (DCM). We sought to probe these effects further using population genomics and in vivo murine models. We crossed Mtss1-/- mice with a transgenic (Tg) murine model of human DCM caused by the D230N pathogenic mutation in Tpm1 (tropomyosin 1). In females, Tg/Mtss1+/- mice had significantly increased LV ejection fraction and reduced LV volumes relative to their Tg/Mtss1+/+ counterparts, signifying partial rescue of DCM due to Mtss1 haploinsufficiency. No differences were observed in males. To study effects in humans, we fine-mapped the MTSS1 locus with 82 cardiac magnetic resonance (CMR) traits in 22,381 UK Biobank participants. MTSS1 enhancer variants showed interaction with biological sex in their associations with several CMR traits. After stratification by biological sex, associations with CMR traits and colocalization with MTSS1 expression in the Genotype-Tissue Expression (GTEx) Project were observed principally in women and were substantially weaker in men. These findings suggest sex dimorphism in the effects of MTSS1-lowering alleles, and parallel the increased LV ejection fraction and reduced LV volumes observed female Tg/Mtss1+/- mice. Together, our findings at the MTSS1 locus suggest a genetic basis for sex dimorphism in cardiac remodeling and motivate sex-specific study of common variants associated with cardiac traits and disease.

A survey of recent methods for addressing AI fairness and bias in biomedicine.

Objectives: Artificial intelligence (AI) systems have the potential to revolutionize clinical practices, including improving diagnostic accuracy and surgical decision-making, while also reducing costs and manpower. However, it is important to recognize that these systems may perpetuate social inequities or demonstrate biases, such as those based on race or gender. Such biases can occur before, during, or after the development of AI models, making it critical to understand and address potential biases to enable the accurate and reliable application of AI models in clinical settings. To mitigate bias concerns during model development, we surveyed recent publications on different debiasing methods in the fields of biomedical natural language processing (NLP) or computer vision (CV). Then we discussed the methods, such as data perturbation and adversarial learning, that have been applied in the biomedical domain to address bias. Methods: We performed our literature search on PubMed, ACM digital library, and IEEE Xplore of relevant articles published between January 2018 and December 2023 using multiple combinations of keywords. We then filtered the result of 10,041 articles automatically with loose constraints, and manually inspected the abstracts of the remaining 890 articles to identify the 55 articles included in this review. Additional articles in the references are also included in this review. We discuss each method and compare its strengths and weaknesses. Finally, we review other potential methods from the general domain that could be applied to biomedicine to address bias and improve fairness. Results: The bias of AIs in biomedicine can originate from multiple sources such as insufficient data, sampling bias and the use of health-irrelevant features or race-adjusted algorithms. Existing debiasing methods that focus on algorithms can be categorized into distributional or algorithmic. Distributional methods include data augmentation, data perturbation, data reweighting methods, and federated learning. Algorithmic approaches include unsupervised representation learning, adversarial learning, disentangled representation learning, loss-based methods and causality-based methods.

GeneGPT: augmenting large language models with domain tools for improved access to biomedical information.

MOTIVATION: While large language models (LLMs) have been successfully applied to various tasks, they still face challenges with hallucinations. Augmenting LLMs with domain-specific tools such as database utilities can facilitate easier and more precise access to specialized knowledge. In this article, we present GeneGPT, a novel method for teaching LLMs to use the Web APIs of the National Center for Biotechnology Information (NCBI) for answering genomics questions. Specifically, we prompt Codex to solve the GeneTuring tests with NCBI Web APIs by in-context learning and an augmented decoding algorithm that can detect and execute API calls. RESULTS: Experimental results show that GeneGPT achieves state-of-the-art performance on eight tasks in the GeneTuring benchmark with an average score of 0.83, largely surpassing retrieval-augmented LLMs such as the new Bing (0.44), biomedical LLMs such as BioMedLM (0.08) and BioGPT (0.04), as well as GPT-3 (0.16) and ChatGPT (0.12). Our further analyses suggest that: First, API demonstrations have good cross-task generalizability and are more useful than documentations for in-context learning; second, GeneGPT can generalize to longer chains of API calls and answer multi-hop questions in GeneHop, a novel dataset introduced in this work; finally, different types of errors are enriched in different tasks, providing valuable insights for future improvements. AVAILABILITY AND IMPLEMENTATION: The GeneGPT code and data are publicly available at https://github.com/ncbi/GeneGPT.