SPeC: A Soft Prompt-Based Calibration on Performance Variability of Large Language Model in Clinical Notes Summarization.

Electronic health records (EHRs) store an extensive array of patient information, encompassing medical histories, diagnoses, treatments, and test outcomes. These records are crucial for enabling healthcare providers to make well-informed decisions regarding patient care. Summarizing clinical notes further assists healthcare professionals in pinpointing potential health risks and making better-informed decisions. This process contributes to reducing errors and enhancing patient outcomes by ensuring providers have access to the most pertinent and current patient data. Recent research has shown that incorporating instruction prompts with large language models (LLMs) substantially boosts the efficacy of summarization tasks. However, we show that this approach also leads to increased performance variance, resulting in significantly distinct summaries even when instruction prompts share similar meanings. To tackle this challenge, we introduce a model-agnostic Soft Prompt-BasedCalibration (SPeC) pipeline that employs soft prompts to lower variance while preserving the advantages of prompt-based summarization. Experimental findings on multiple clinical note tasks and LLMs indicate that our method not only bolsters performance but also effectively regulates variance across different LLMs, providing a more consistent and reliable approach to summarizing critical medical information.

A Comparative Study of Structural Deformation Test Based on Edge Detection and Digital Image Correlation.

Digital image-correlation (DIC) algorithms rely heavily on the accuracy of the initial values provided by whole-pixel search algorithms for structural displacement monitoring. When the measured displacement is too large or exceeds the search domain, the calculation time and memory consumption of the DIC algorithm will increase greatly, and even fail to obtain the correct result. The paper introduced two edge-detection algorithms, Canny and Zernike moments in digital image-processing (DIP) technology, to perform geometric fitting and sub-pixel positioning on the specific pattern target pasted on the measurement position, and to obtain the structural displacement according to the change of the target position before and after deformation. This paper compared the difference between edge detection and DIC in accuracy and calculation speed through numerical simulation, laboratory, and field tests. The study demonstrated that the structural displacement test based on edge detection is slightly inferior to the DIC algorithm in terms of accuracy and stability. As the search domain of the DIC algorithm becomes larger, its calculation speed decreases sharply, and is obviously slower than the Canny and Zernike moment algorithms.

Transcriptomics and metagenomics of common cutworm (Spodoptera litura) and fall armyworm (Spodoptera frugiperda) demonstrate differences in detoxification and development.

BACKGROUND: Spodoptera litura is an important polyphagous pest that causes significant damage to the agricultural sector. We performed RNA-seq of 15 S. litura individuals from larval (fifth and sixth instar larvae), chrysalis, and adult developmental stages. We also compared the S. litura transcriptome data with Spodoptera frugiperda across the same developmental stages, which was sequenced in our previous study. RESULTS: A total of 101,885 differentially expressed transcripts (DETs) were identified in S. litura. Gene Ontology (GO) and Kyoto Encyclopedia of Genes and Genomes (KEGG) enrichment analyses indicated that S. litura may undergo active xenobiotic and detoxifying metabolism during its larval and adult stages, which may explain difficulties with current population control measures. We also found that DETs of single-copy orthologous genes between S. litura and S. frugiperda were involved in basic metabolism and development. However, energy and metabolic processes genes had a higher expression in S. litura, whereas nervous and olfactory function genes had a higher expression in S. frugiperda. Metagenomics analysis in larval S. litura and S. frugiperda revealed that microbiota participate in the detoxification and metabolism processes, but the relative abundance of detoxification-related microbiota was more abundant in S. frugiperda. Transcriptome results also confirmed the detoxification-related pathway of S. frugiperda was more abundant than in S. litura. CONCLUSIONS: Significant changes at transcriptional level were identified during the different development stages of S. litura. Importantly, we also identified detoxification associated genes and gut microbiota between S. litura and S. frugiperda at different developmental stages, which will be valuable in revealing possible mechanisms of detoxification and development in these two lepidopterans.

Sex-specific gene expression in the blood of four primates.

Blood is an important non-reproductive tissue, but little is known about the sex-specific gene expressions in the blood. Therefore, we investigated sex-specific gene expression differences in the blood tissues of four primates, rhesus macaques (Macaca mulatta), Tibetan macaques (M. thibetana), yellow baboons (Papio cynocephalus), and humans. We identified seven sex-specific differentially expressed genes (SDEGs) in each non-human primate and 31 SDEGs in humans. The four primates had only one common SDEG, MAP7D2. In humans, immune-related SDEGs were identified as up-regulated, but also down-regulated in females. We also found that most of the X-Y gene pairs had similar expression levels between species, except pair EIF1AY/EIF1AX. The expression level of X-Y gene pairs of rhesus and Tibetan macaques showed no significant differential expression levels, while humans had six significant XY-biased and three XX-biased X-Y gene pairs. Our observed sex differences in blood should increase understanding of sex differences in primate blood tissue.

Age-related gene expression and DNA methylation changes in rhesus macaque.

Aging is a very complicated biological process that can change gene expressions. The Chinese rhesus macaque (Macaca mulatta lasiota; CR) is closely related to humans. We explored gene expression with increasing age and DNA methylation changes in young and old CRs. Results showed blood transcriptome and DNA methylome significantly changed from young to old CRs. The age-associated differentially expressed genes (DEGs) and differentially methylated regions (DMRs) were associated with age-related biological features, such as immunity, blood coagulation, and biosynthetic process. The measurements of coagulation indicators confirmed old CRs had shorter coagulation time than young CRs, and the activities of coagulation factor II (FII) and factor VIII (FVIII) were enhanced in old CRs. Humans and CRs exhibited the same enhanced blood coagulation with age phenotype. Our study found aging is a critical factor affecting gene expression in CRs, and also provided new insights into the blood coagulation changes in non-human primates.

Comprehensive analysis of lncRNA and mRNA expression changes in Tibetan chicken lung tissue between three developmental stages.

The Tibetan chicken is a native Chinese breed that lives at high elevations and has adapted to the extreme environmental conditions of the Tibetan Plateau. However, its hypoxic adaptation at the gene expression level is unclear. Here, we sequenced nine lung transcriptomes of the Tibetan chicken at three developmental stages (5 and 42 weeks and 4.5 years). A total of 1.02 billion clean reads were obtained. We identified 16 012 mRNAs and 6898 lncRNAs. The expression of mRNA showed that nine samples were significantly divided into three clusters, with higher correlation and closer relationship between the 5 and 42 week groups. We identified 399 differentially expressed genes (DEGs) between the 5 and 42 week groups, 3532 DEGs between the 5 week and 4.5 year groups, and 3909 DEGs between the 42 week and 4.5 year groups. The up-regulated DEGs in the 5 week group, compared with 42 week and 4.5 year groups, were enriched in GO terms associated with growth and development, whereas the up-regulated DEGs in the 4.5 year group were mainly enriched in many metabolic-related categories. Moreover, the enrichment results with up-regulated DEGs in the 5 and/or 42 week groups, compared with the 4.5 year group, were associated with hypoxic adaptation, such as oxygen transport, oxygen binding and oxygen carrier activity, and calcium signaling pathway. In addition, we identified 978 high-correlation lncRNA and protein-coding gene pairs, and 524 significant neighboring protein-coding genes were also DEGs. Our results provide new insights into gene expression of lung tissue in Tibetan chickens during the aging process.

TSPAN12 promotes chemoresistance and proliferation of SCLC under the regulation of miR-495.

Studies have shown that TSPAN12 serves as an oncogenic gene or tumor suppressor depending on the types of cancer. However, its role in the drug resistance of small cell lung cancer (SCLC) is still unknown. In this study, we investigated whether TSPAN12 regulates chemoresistance, proliferation and tumor growth of SCLC under the regulation of miR-495 using two drug resistant cell lines. The results showed that down-regulation of TSPAN12 inhibited cells chemoresistance, proliferation and tumor growth. Besides, TSPAN12 elevation in SCLC specimens correlated with poor pathologic stage and shorter survival time. In addition, the dual luciferase assay indicated that TSPAN12 was one of the directly targeted genes of miR-495. Our study revealed that TSPAN12 promoted chemoresistance of SCLC under the regulation of miR-495. TSPAN12 depletion is a promising strategy for inhibiting the chemoresistance in SCLC.

Long noncoding RNA-HOTAIR affects chemoresistance by regulating HOXA1 methylation in small cell lung cancer cells.

Homeobox (HOX) transcript antisense RNA (HOTAIR), a long intergenic noncoding RNA (lincRNA), has been reported to play an oncogenic role in various cancers including small cell lung cancer (SCLC). However, it is not known whether HOTAIR can modulate chemoresistance in SCLC. The aim of this study is to investigate the roles of HOTAIR in chemoresistance of SCLC and its possible molecular mechanism. Knockdown of HOTAIR was carried out in SCLC multidrug-resistant cell lines (H69AR and H446AR) and the parental cell lines (H69 and H446) to assess its influence on chemoresistance. The results showed that downregulation of HOTAIR increased cell sensitivity to anticancer drugs through increasing cell apoptosis and cell cycle arrest, and suppressed tumor growth in vivo. Moreover, HOXA1 methylation increased in the resistant cells using bisulfite sequencing PCR. Depletion of HOTAIR reduced HOXA1 methylation by decreasing DNMT1 and DNMT3b expression. The interaction between HOTAIR and HOXA1 was validated by RNA immunoprecipitation. Taken together, our study suggested that HOTAIR mediates chemoresistance of SCLC by regulating HOXA1 methylation and could be utilized as a potential target for new adjuvant therapies against chemoresistance.

WW domain binding protein 5 induces multidrug resistance of small cell lung cancer under the regulation of miR-335 through the Hippo pathway.

BACKGROUND: Our previous study indicated that WW domain binding protein 5 (WBP5) expression was elevated significantly in a drug-resistant cell compared with its parental cell. Nevertheless, its functional role and underlying mechanisms remain unknown. METHODS: In this study, WBP5 was examined in 62 small cell lung cancer (SCLC) patient samples by immunohistochemical technique. Stable WBP5-overexpressed and WBP5-underexpressed cells were further established to assess the role of WBP5 in drug resistance, apoptosis and tumour growth. We also conducted western blot to detect the expression of MST2 and YAP1 and their phosphorylated protein. RESULTS: The results revealed that WBP5 expression was significantly associated with the shorter survival time in SCLC patients. Upregulation of WBP5 induced multidrug resistance (MDR) and decreased apoptosis, whereas downregulation of WBP5 enhanced drug sensitivity and increased apoptosis. We also found that miR-335 negatively regulated the MDR of WBP5 by targeting its 3'UTR. Furthermore, WBP5 can lower YAP1 phosphorylation at Serine 127 and induce nuclear accumulation of YAP1. Inhibition of YAP1 by Verteporfin could blunt the MDR phenotype of WBP5. CONCLUSIONS: WW domain binding protein 5 can modulate MDR through the Hippo pathway under the regulation of miR-335. WW domain binding protein 5 may be a prognostic predictor and a potential target for interfering with MDR in SCLC.

Zinc finger E-box-binding homeobox 2 (ZEB2) regulated by miR-200b contributes to multi-drug resistance of small cell lung cancer.

Zinc finger E-box-binding homeobox 2 (ZEB2) was closely related to the oncogenesis, development and response to chemotherapy of cancer. However, its biological functions in small cell lung cancer (SCLC) remain unknown. The aim of this study is to investigate the roles of ZEB2 in chemoresistance of SCLC and its possible molecular mechanism. Expression of ZEB2 was examined in sixty-eight cases of SCLC tissues by immunohistochemistry. Knockdown of ZEB2 was carried out in SCLC multidrug resistant cells (H69AR) to assess its influence on chemoresistance. The results showed that ZEB2 was expressed in 23.5% (16/68) of SCLC. Overexpression of ZEB2 was associated with the poor pathologic stage of SCLC (P < 0.001 by the Fisher's Exact Test) and the shorter survival time (by the Kaplan-Meier method). Inhibition of ZEB2 expression using small interfering RNA in H69AR cells sensitized cancer cells to chemotherapeutic drugs through increasing drug-induced cell apoptosis accompanied with S phase arrest. In silico analysis demonstrated that there are complementary binding sites between miR-200b and ZEB2 3'-UTR, and identified miR-200b as a potential regulator of ZEB2. We found that miR-200b was down-regulated in the resistant cells and enforced expression of miR-200b by miRNA mimics increased cell sensitivity. Overexpression of miR-200b led to the downregulation of ZEB2 at protein level. Luciferase reporter gene assay showed that 3'UTR ZEB2 activity was regulated by miR-200b. Our results suggest that ZEB2 modulates drug resistance and is regulated by miR-200b. All findings provide insight into the ZEB2 signaling mechanism and ZEB2 may be a potentially novel target for multi-drug resistance in SCLC.

Large Language Models for Healthcare Data Augmentation: An Example on Patient-Trial Matching.

The process of matching patients with suitable clinical trials is essential for advancing medical research and providing optimal care. However, current approaches face challenges such as data standardization, ethical considerations, and a lack of interoperability between Electronic Health Records (EHRs) and clinical trial criteria. In this paper, we explore the potential of large language models (LLMs) to address these challenges by leveraging their advanced natural language generation capabilities to improve compatibility between EHRs and clinical trial descriptions. We propose an innovative privacy-aware data augmentation approach for LLM-based patient-trial matching (LLM-PTM), which balances the benefits of LLMs while ensuring the security and confidentiality of sensitive patient data. Our experiments demonstrate a 7.32% average improvement in performance using the proposed LLM-PTM method, and the generalizability to new data is improved by 12.12%. Additionally, we present case studies to further illustrate the effectiveness of our approach and provide a deeper understanding of its underlying principles.

Local Contrastive Learning for Medical Image Recognition.

The proliferation of Deep Learning (DL)-based methods for radiographic image analysis has created a great demand for expert-labeled radiology data. Recent self-supervised frameworks have alleviated the need for expert labeling by obtaining supervision from associated radiology reports. These frameworks, however, struggle to distinguish the subtle differences between different pathologies in medical images. Additionally, many of them do not provide interpretation between image regions and text, making it difficult for radiologists to assess model predictions. In this work, we propose Local Region Contrastive Learning (LRCLR), a flexible fine-tuning framework that adds layers for significant image region selection as well as cross-modality interaction. Our results on an external validation set of chest x-rays suggest that LRCLR identifies significant local image regions and provides meaningful interpretation against radiology text while improving zero-shot performance on several chest x-ray medical findings.

Analysis of coexisting pathogens in nasopharyngeal swabs from COVID-19.

Background: The impact of COVID-19 on the world is still ongoing, and it is currently under regular management. Although most infected people have flu-like symptoms and can cure themselves, coexisting pathogens in COVID-19 patients should not be taken lightly. The present study sought to investigate the coexisting pathogens in SARS-CoV-2 infected patients and identify the variety and abundance of dangerous microbes to guide treatment strategies with a better understanding of the untested factors. Methods: We extracted total DNA and RNA in COVID-19 patient specimens from nasopharyngeal swabs to construct a metagenomic library and utilize Next Generation Sequencing (NGS) to discover chief bacteria, fungi, and viruses in the body of patients. High-throughput sequencing data from Illumina Hiseq 4000 were analyzed using Krona taxonomic methodology for species diversity. Results: We studied 56 samples to detect SARS-CoV-2 and other pathogens and analyzed the species diversity and community composition of these samples after sequencing. Our results showed some threatening pathogens such as Mycoplasma pneumoniae, Klebsiella pneumoniae, Streptococcus pneumoniae, and some previously reported pathogens. SARS-CoV-2 combined with bacterial infection is more common. The results of heat map analysis showed that the abundance of bacteria was mostly more than 1000 and that of viruses was generally less than 500. The pathogens most likely to cause SARS-CoV-2 coinfection or superinfection include Streptococcus pneumoniae, Haemophilus influenzae, Staphylococcus aureus, Klebsiella pneumoniae, and Human gammaherpesvirus 4. Conclusions: The current coinfection and superinfection status is not optimistic. Bacteria are the major threat group that increases the risk of complications and death in COVID-19 patients and attention should be paid to the use and control of antibiotics. Our study investigated the main types of respiratory pathogens prone to coexisting or superinfection in COVID-19 patients, which is valuable for identifying and treating SARS-CoV-2.

DiscoverPath: A Knowledge Refinement and Retrieval System for Interdisciplinarity on Biomedical Research.

The exponential growth in scholarly publications necessitates advanced tools for efficient article retrieval, especially in interdisciplinary fields where diverse terminologies are used to describe similar research. Traditional keyword-based search engines often fall short in assisting users who may not be familiar with specific terminologies. To address this, we present a knowledge graph based paper search engine for biomedical research to enhance the user experience in discovering relevant queries and articles. The system, dubbed DiscoverPath, employs Named Entity Recognition (NER) and part-of-speech (POS) tagging to extract terminologies and relationships from article abstracts to create a KG. To reduce information overload, DiscoverPath presents users with a focused subgraph containing the queried entity and its neighboring nodes and incorporates a query recommendation system enabling users to iteratively refine their queries. The system is equipped with an accessible Graphical User Interface that provides an intuitive visualization of the KG, query recommendations, and detailed article information, enabling efficient article retrieval, thus fostering interdisciplinary knowledge exploration. DiscoverPath is open-sourced at https://github.com/ynchuang/DiscoverPath with a demo video at Youtube.

Defense Against Explanation Manipulation.

Explainable machine learning attracts increasing attention as it improves the transparency of models, which is helpful for machine learning to be trusted in real applications. However, explanation methods have recently been demonstrated to be vulnerable to manipulation, where we can easily change a model's explanation while keeping its prediction constant. To tackle this problem, some efforts have been paid to use more stable explanation methods or to change model configurations. In this work, we tackle the problem from the training perspective, and propose a new training scheme called Adversarial Training on EXplanations (ATEX) to improve the internal explanation stability of a model regardless of the specific explanation method being applied. Instead of directly specifying explanation values over data instances, ATEX only puts constraints on model predictions which avoids involving second-order derivatives in optimization. As a further discussion, we also find that explanation stability is closely related to another property of the model, i.e., the risk of being exposed to adversarial attack. Through experiments, besides showing that ATEX improves model robustness against manipulation targeting explanation, it also brings additional benefits including smoothing explanations and improving the efficacy of adversarial training if applied to the model.

Fairly Predicting Graft Failure in Liver Transplant for Organ Assigning.

Liver transplant is an essential therapy performed for severe liver diseases. The fact of scarce liver resources makes the organ assigning crucial. Model for End-stage Liver Disease (MELD) score is a widely adopted criterion when making organ distribution decisions. However, it ignores post-transplant outcomes and organ/donor features. These limitations motivate the emergence of machine learning (ML) models. Unfortunately, ML models could be unfair and trigger bias against certain groups of people. To tackle this problem, this work proposes a fair machine learning framework targeting graft failure prediction in liver transplant. Specifically, knowledge distillation is employed to handle dense and sparse features by combining the advantages of tree models and neural networks. A two-step debiasing method is tailored for this framework to enhance fairness. Experiments are conducted to analyze unfairness issues in existing models and demonstrate the superiority of our method in both prediction and fairness performance.

Comparative transcriptomes of nine tissues for the Heilongjiang brown frog (Rana amurensis).

The Heilongjiang brown frog (Rana amurensis) is widely used in traditional Chinese medicine. In particular, the oviduct and skin have been developed into various health products. However, limited numbers of complete genomes of amphibian species have been reported, excluding the Heilongjiang brown frog. Here, the transcriptomes of 45 samples from the liver, spleen, heart, ovaries, thigh muscles, skin, oviduct, stomach and intestine of five Heilongjiang brown frog were reassembled and analyzed. A total of 1,085,532 unigenes with an average length of 676.6 bp and N50 of 722 bp were obtained. Comparative transcriptomics of different tissues detected tissue-specific expression. There were 3248 differentially expressed genes (DEGs) in the ovary, and the number of unique DEGs between the ovary and spleen was the largest. The results of DEGs enrichment showed there were many pathways and items related to protein synthesis and metabolism in the oviduct. The DEGs of the skin were enriched with many bacterial defense items, indicating that there were a large number of antimicrobial peptides in the skin. Thus, these were suitable as biological sources for the development and extraction of antimicrobial peptides. Through the assembly of transcriptome sequencing data and functional annotation of the Heilongjiang brown frog genome, this study provides reference materials for further exploring and utilizing functional gene resources of frogs and lays a foundation for medical research and the development of new products.

Comparative transcriptomes of three different skin sites for the Asiatic toad (Bufo gargarizans).

Toads release toxic dry secretions from glands in their skin. Toxin possesses a wide range of biological effects, but little is known about its specific gene expression pattern and regulatory mechanisms. The Asiatic toad (Bufo gargarizans) is widely used to produce toxin. Here, we explored the gene expression of 30 tissue samples from three different skin sites (parotoid gland, dorsal skin, and abdomen skin) of B. gargarizans. After de novo assembly, 783,130 unigenes with an average length of 489 bp (N50 = 556 bp) were obtained. A total of 9,248 significant differentially expressed genes (DEGs) were detected. There were 8,819 DEGs between the parotoid gland and abdomen skin and 1,299 DEGs between the dorsal skin and abdomen skin, while only 1,283 DEGs were obtained between the parotoid gland and dorsal skin. Through enrichment analysis, it was found that the detected differential gene expressions corresponded to the different functions of different skin sites. Our key findings were the genetic expression of toxin secretion, the protection function of skin, and the related genes such as HSD3B, Cyp2c, and CAT, LGALS9. In conclusion, we provide useful transcript resources to study the gene expression and gene function of B. gargarizans and other amphibians. The detected DEGs between different sites of the skin provided better insights into the genetic mechanisms of toxin secretion and the protection function of skin for amphibians.

Two chronically misdiagnosed patients infected with Nocardia cyriacigeorgica accurately diagnosed by whole genome resequencing.

Nocardiosis is a rare but life-threatening infection particularly affecting immuno-compromised hosts, causing localized or systemic suppurative disease usually in human beings. Nocardia species, as the pathogen of nocardiosis, are difficult to differentiate because of their complex colony morphological features. In this study, we describe two patients who had been misdiagnosed for a long time infected with Nocardia cyriacigeorgica with completely different morphology were accurately diagnosed. Single colonies were analyzed by Gram staining, acid-fast stain, mass spectrometry and whole genome resequencing (WGRS). These two bacterial, strains L5.53 and L5.54, were found to be Gram-negative and acid-fast-weak positive. Blood sample culturing of strain L5.53 yielded white colonies, which were like a layer of hoarfrost, while colonies of L5.54 were yellow, rough, slightly convex. The two strains were identified as Nocardia sp. by mass spectrometry, and WGRS accurately determined them as N. cyriacigeorgica. After medical treatment, one patient was cured and the other was still receiving treatment in the hospital. It can be seen that Nocardia sp. cannot be accurately classified and identified only by phenotypic tests such as bacterial morphological differences, so it is necessary to identify Nocardia spp. with phenotypic tests in combination with other molecular biology technologies, such as WGRS.