Natural language processing to identify lupus nephritis phenotype in electronic health records

BACKGROUND

Systemic lupus erythematosus (SLE) is a rare autoimmune disorder characterized by an unpredictable course of flares and remission with diverse manifestations. Lupus nephritis, one of the major disease manifestations of SLE for organ damage and mortality, is a key component of lupus classification criteria. Accurately identifying lupus nephritis in electronic health records (EHRs) would therefore benefit large cohort observational studies and clinical trials where characterization of the patient population is critical for recruitment, study design, and analysis. Lupus nephritis can be recognized through procedure codes and structured data, such as laboratory tests. However, other critical information documenting lupus nephritis, such as histologic reports from kidney biopsies and prior medical history narratives, require sophisticated text processing to mine information from pathology reports and clinical notes. In this study, we developed algorithms to identify lupus nephritis with and without natural language processing (NLP) using EHR data from the Northwestern Medicine Enterprise Data Warehouse (NMEDW).

METHODS

We developed five algorithms: a rule-based algorithm using only structured data (baseline algorithm) and four algorithms using different NLP models. The first NLP model applied simple regular expression for keywords search combined with structured data. The other three NLP models were based on regularized logistic regression and used different sets of features including positive mention of concept unique identifiers (CUIs), number of appearances of CUIs, and a mixture of three components (i.e. a curated list of CUIs, regular expression concepts, structured data) respectively. The baseline algorithm and the best performing NLP algorithm were externally validated on a dataset from Vanderbilt University Medical Center (VUMC).

RESULTS

Our best performing NLP model incorporated features from both structured data, regular expression concepts, and mapped concept unique identifiers (CUIs) and showed improved F measure in both the NMEDW (0.41 vs 0.79) and VUMC (0.52 vs 0.93) datasets compared to the baseline lupus nephritis algorithm.

CONCLUSION

Our NLP MetaMap mixed model improved the F-measure greatly compared to the structured data only algorithm in both internal and external validation datasets. The NLP algorithms can serve as powerful tools to accurately identify lupus nephritis phenotype in EHR for clinical research and better targeted therapies.

[Association of lifestyle and apolipoprotein E gene with risk for cognitive frailty in elderly population in China]

Objective: To investigate the impact of lifestyle, apolipoprotein E (ApoE) gene, and their interaction on the risk for cognitive frailty in the elderly population in China. Methods: The study participants were from the Chinese Longitudinal Healthy Longevity Survey. The information about their lifestyles were collected by questionnaire survey, and a weighted lifestyle score was constructed based on β coefficients associated with specific lifestyles to assess the combined lifestyle. ApoE genotypes were assessed by rs429358 and rs7412 single nucleotide polymorphisms. Cognitive frailty was assessed based on cognitive function and physical frailty. Cox proportional hazards regression model was used to analyze the association of lifestyle and ApoE gene with the risk for cognitive frailty and evaluate the multiplicative and additive interactions between lifestyle and ApoE gene. Results: A total of 5 676 elderly persons, with median age [M (Q1, Q3)] of 76 (68, 85) years, were included, in whom 615 had cognitive frailty. The analysis by Cox proportional hazards regression model indicated that moderate and high levels of dietary diversity could reduce the risk for cognitive frailty by 18% [hazard ratio (HR)=0.82, 95%CI: 0.68-1.00] and 28% (HR=0.72, 95%CI: 0.57-0.91), respectively; moderate and high levels of physical activity could reduce the risk by 31% (HR=0.69, 95%CI: 0.56-0.85) and 23% (HR=0.77, 95%CI: 0.64-0.93), respectively. Healthy lifestyle was associated with a 40% reduced risk for cognitive frailty (HR=0.60, 95%CI: 0.46-0.78). ApoE ε4 allele was associated with a 26% increased risk for cognitive frailty (HR=1.26, 95%CI: 1.02-1.56). No multiplicative or additive interactions were found between lifestyle and ApoE gene. Conclusions: Dietary diversity and regular physical activity have protective effects against cognitive frailty in elderly population. Healthy lifestyle can reduce the risk for cognitive frailty in elderly population regardless of ApoE ε4 allele carriage status.

AD-BERT: Using pre-trained language model to predict the progression from mild cognitive impairment to Alzheimer's disease

OBJECTIVE

We develop a deep learning framework based on the pre-trained Bidirectional Encoder Representations from Transformers (BERT) model using unstructured clinical notes from electronic health records (EHRs) to predict the risk of disease progression from Mild Cognitive Impairment (MCI) to Alzheimer's Disease (AD).

METHODS

We identified 3657 patients diagnosed with MCI together with their progress notes from Northwestern Medicine Enterprise Data Warehouse (NMEDW) between 2000 and 2020. The progress notes no later than the first MCI diagnosis were used for the prediction. We first preprocessed the notes by deidentification, cleaning and splitting into sections, and then pre-trained a BERT model for AD (named AD-BERT) based on the publicly available Bio+Clinical BERT on the preprocessed notes. All sections of a patient were embedded into a vector representation by AD-BERT and then combined by global MaxPooling and a fully connected network to compute the probability of MCI-to-AD progression. For validation, we conducted a similar set of experiments on 2563 MCI patients identified at Weill Cornell Medicine (WCM) during the same timeframe.

RESULTS

Compared with the 7 baseline models, the AD-BERT model achieved the best performance on both datasets, with Area Under receiver operating characteristic Curve (AUC) of 0.849 and F1 score of 0.440 on NMEDW dataset, and AUC of 0.883 and F1 score of 0.680 on WCM dataset.

CONCLUSION

The use of EHRs for AD-related research is promising, and AD-BERT shows superior predictive performance in modeling MCI-to-AD progression prediction. Our study demonstrates the utility of pre-trained language models and clinical notes in predicting MCI-to-AD progression, which could have important implications for improving early detection and intervention for AD.

Predicting Recurrence in Pancreatic Ductal Adenocarcinoma after Radical Surgery Using an AX-Unet Pancreas Segmentation Model and Dynamic Nomogram

This study aims to investigate the reliability of radiomic features extracted from contrast-enhanced computer tomography (CT) by AX-Unet, a pancreas segmentation model, to analyse the recurrence of pancreatic ductal adenocarcinoma (PDAC) after radical surgery. In this study, we trained an AX-Unet model to extract the radiomic features from preoperative contrast-enhanced CT images on a training set of 205 PDAC patients. Then we evaluated the segmentation ability of AX-Unet and the relationship between radiomic features and clinical characteristics on an independent testing set of 64 patients with clear prognoses. The lasso regression analysis was used to screen for variables of interest affecting patients' post-operative recurrence, and the Cox proportional risk model regression analysis was used to screen for risk factors and create a nomogram prediction model. The proposed model achieved an accuracy of 85.9% for pancreas segmentation, meeting the requirements of most clinical applications. Radiomic features were found to be significantly correlated with clinical characteristics such as lymph node metastasis, resectability status, and abnormally elevated serum carbohydrate antigen 19-9 (CA 19-9) levels. Specifically, variance and entropy were associated with the recurrence rate (p < 0.05). The AUC for the nomogram predicting whether the patient recurred after surgery was 0.92 (95% CI: 0.78-0.99) and the C index was 0.62 (95% CI: 0.48-0.78). The AX-Unet pancreas segmentation model shows promise in analysing recurrence risk factors after radical surgery for PDAC. Additionally, our findings suggest that a dynamic nomogram model based on AX-Unet can provide pancreatic oncologists with more accurate prognostic assessments for their patients.

Unique molecular signatures sustained in circulating monocytes and regulatory T cells in convalescent COVID-19 patients

Over two years into the COVID-19 pandemic, the human immune response to SARS-CoV-2 during the active disease phase has been extensively studied. However, the long-term impact after recovery, which is critical to advance our understanding SARS-CoV-2 and COVID-19-associated long-term complications, remains largely unknown. Herein, we characterized single-cell profiles of circulating immune cells in the peripheral blood of 100 patients, including convalescent COVID-19 and sero-negative controls. Flow cytometry analyses revealed reduced frequencies of both short-lived monocytes and long-lived regulatory T (Treg) cells within the patients who have recovered from severe COVID-19. sc-RNA seq analysis identifies seven heterogeneous clusters of monocytes and nine Treg clusters featuring distinct molecular signatures in association with COVID-19 severity. Asymptomatic patients contain the most abundant clusters of monocytes and Tregs expressing high CD74 or IFN-responsive genes. In contrast, the patients recovered from a severe disease have shown two dominant inflammatory monocyte clusters featuring S100 family genes: one monocyte cluster of S100A8 & A9 coupled with high HLA-I and another cluster of S100A4 & A6 with high HLA-II genes, a specific non-classical monocyte cluster with distinct IFITM family genes, as well as a unique TGF-β high Treg Cluster. The outpatients and seronegative controls share most of the monocyte and Treg clusters patterns with high expression of HLA genes. Surprisingly, while presumably short-lived monocytes appear to have sustained alterations over 4 months, the decreased frequencies of long-lived Tregs (high HLA-DRA and S100A6) in the outpatients restore over the tested convalescent time (≥ 4 months). Collectively, our study identifies sustained and dynamically altered monocytes and Treg clusters with distinct molecular signatures after recovery, associated with COVID-19 severity.

Neck management of pathological N1 oral squamous cell carcinoma: a retrospective study

This study was performed to compare the effects of neck dissection procedures on the prognosis of patients with pathological N1 (pN1) oral squamous cell carcinoma (OSCC), analyse factors affecting the prognosis, and provide a neck management strategy for clinical N1 (cN1) oral cancer. The study patients were divided into two groups according to the neck dissection: a selective neck dissection (SND) group (n = 85) and a radical or modified radical neck dissection (RND/MRND) group (n = 22). There was no statistically significant difference in recurrence rates at local, regional, and distant sites between the SND and RND/MRND groups. The 5-year overall survival was 68.3% for SND and 65.2% for RND/MRND patients (P = 0.590), while the 5-year disease-specific survival was 70.4% for SND and 75.7% for RND/MRND patients (P = 0.715). Histological grade and postoperative radiotherapy were independent predictors of the outcome for SND patients. For histological grade II/III cases, 5-year overall survival (P = 0.004) and disease-specific survival (P = 0.002) outcomes differed significantly between patients treated with and without postoperative radiotherapy, with worse survival for patients not treated with radiotherapy. Therefore, SND appears appropriate for cN1 OSCC patients, and postoperative radiotherapy is recommended for those with histological grade II or III tumours.

Comparing the effects of four common drug classes on the progression of mild cognitive impairment to dementia using electronic health records

The objective of this study was to investigate the potential association between the use of four frequently prescribed drug classes, namely antihypertensive drugs, statins, selective serotonin reuptake inhibitors, and proton-pump inhibitors, and the likelihood of disease progression from mild cognitive impairment (MCI) to dementia using electronic health records (EHRs). We conducted a retrospective cohort study using observational EHRs from a cohort of approximately 2 million patients seen at a large, multi-specialty urban academic medical center in New York City, USA between 2008 and 2020 to automatically emulate the randomized controlled trials. For each drug class, two exposure groups were identified based on the prescription orders documented in the EHRs following their MCI diagnosis. During follow-up, we measured drug efficacy based on the incidence of dementia and estimated the average treatment effect (ATE) of various drugs. To ensure the robustness of our findings, we confirmed the ATE estimates via bootstrapping and presented associated 95% confidence intervals (CIs). Our analysis identified 14,269 MCI patients, among whom 2501 (17.5%) progressed to dementia. Using average treatment estimation and bootstrapping confirmation, we observed that drugs including rosuvastatin (ATE = - 0.0140 [- 0.0191, - 0.0088], p value < 0.001), citalopram (ATE = - 0.1128 [- 0.125, - 0.1005], p value < 0.001), escitalopram (ATE = - 0.0560 [- 0.0615, - 0.0506], p value < 0.001), and omeprazole (ATE = - 0.0201 [- 0.0299, - 0.0103], p value < 0.001) have a statistically significant association in slowing the progression from MCI to dementia. The findings from this study support the commonly prescribed drugs in altering the progression from MCI to dementia and warrant further investigation.

Suvemcitug as second-line treatment of advanced or metastatic solid tumors and with FOLFIRI for pretreated metastatic colorectal cancer: phase Ia/Ib open label, dose-escalation trials

BACKGROUND

Suvemcitug (BD0801), a novel humanized rabbit monoclonal antibody against vascular endothelial growth factor, has demonstrated promising antitumor activities in preclinical studies.

PATIENTS AND METHODS

The phase Ia/b trials investigated the safety and tolerability and antitumor activities of suvemcitug for pretreated advanced solid tumors and in combination with FOLFIRI (leucovorin and fluorouracil plus irinotecan) in second-line treatment of metastatic colorectal cancer using a 3 + 3 dose-escalation design. Patients received escalating doses of suvemcitug (phase Ia: 2, 4, 5, 6, and 7.5 mg/kg; phase Ib: 1, 2, 3, 4, and 5 mg/kg plus FOLFIRI). The primary endpoint was safety and tolerability in both trials.

RESULTS

All patients in the phase Ia trial had at least one adverse event (AE). Dose-limiting toxicities included grade 3 hyperbilirubinemia (one patient), hypertension and proteinuria (one patient), and proteinuria (one patient). The maximum tolerated dose was 5 mg/kg. The most common grade 3 and above AEs were proteinuria (9/25, 36%) and hypertension (8/25, 32%). Forty-eight patients (85.7%) in phase Ib had grade 3 and above AEs, including neutropenia (25/56, 44.6%), reduced leucocyte count (12/56, 21.4%), proteinuria (10/56, 17.9%), and elevated blood pressure (9/56, 16.1%). Only 1 patient in the phase Ia trial showed partial response, [objective response rate 4.0%, 95% confidence interval (CI) 0.1% to 20.4%] whereas 18/53 patients in the phase Ib trial exhibited partial response (objective response rate 34.0%, 95% CI 21.5% to 48.3%). The median progression-free survival was 7.2 months (95% CI 5.1-8.7 months).

CONCLUSIONS

Suvemcitug has an acceptable toxicity profile and exhibits antitumor activities in pretreated patients with advanced solid tumors or metastatic colorectal cancer.

[Association of sleep duration and risk of frailty among the elderly over 80 years old in China: a prospective cohort study]

Objective: To explore the association between sleep duration and the risk of frailty among the elderly over 80 years old in China. Methods: Using the data from five surveys of the China Elderly Health Influencing Factors Follow-up Survey (CLHLS) (2005, 2008-2009, 2011-2012, 2014, and 2017-2018), 7 024 elderly people aged 80 years and above were selected as the study subjects. Questionnaires and physical examinations were used to collect information on sleep time, general demographic characteristics, functional status, physical signs, and illness. The frailty state was evaluated based on a frailty index that included 39 variables. The Cox proportional risk regression model was used to analyze the correlation between sleep time and the risk of frailty occurrence. A restricted cubic spline function was used to analyze the dose-response relationship between sleep time and the risk of frailty occurrence. The likelihood ratio test was used to analyze the interaction between age, gender, sleep quality, cognitive impairment, and sleep duration. Results: The age M (Q₁, Q₃) of 7 024 subjects was 87 (82, 92) years old, with a total of 3 435 (48.9%) patients experiencing frailty. The results of restricted cubic spline function analysis showed that there was an approximate U-shaped relationship between sleep time and the risk of frailty. When sleep time was 6.5-8.5 hours, the elderly had the lowest risk of frailty; Multivariate Cox proportional risk regression model analysis showed that compared to 6.5-8.5 hours of sleep, long sleep duration (>8.5 hours) increased the risk of frailty by 13% (HR: 1.13; 95%CI: 1.04-1.22). Conclusion: There is a nonlinear association between sleep time and the risk of frailty in the elderly.

First-in-human, phase I study of AK109, an anti-VEGFR2 antibody in patients with advanced or metastatic solid tumors

BACKGROUND

Vascular endothelial growth factor receptor 2 (VEGFR2) plays a key role in antiangiogenesis which has been an essential strategy for cancer treatment. We report the first-in-human study of AK109, a novel anti-VEGFR2 monoclonal antibody, to characterize the safety profile and pharmacokinetics/pharmacodynamics (PK/PD) properties, and explore the preliminary antitumor efficacy in patients with solid tumors.

PATIENTS AND METHODS

This was a multicenter, open-label, phase I study, including dose escalation and dose expansion (NCT04547205). Patients with advanced cancers were treated 2 and 3 weekly with escalating doses of AK109. A 3 + 3 design was used to determine the maximum tolerated dose. Blood was sampled for PK/PD analysis. The primary endpoint was safety and recommended phase II dose (RP2D).

RESULTS

A total of 40 patients were enrolled. No dose-limiting toxicity was observed. However, 38 patients reported treatment-related adverse events (TRAEs); grade ≥3 TRAEs occurred in 10 patients. The most common TRAEs were proteinuria (n = 24, 60%), hypertension (n = 13, 32.5%), increased aspartate transaminase (n = 11, 27.5%), thrombopenia (n = 10, 25%), and anemia (n = 10, 25%). A total of 28 patients (70%) reported adverse events of special interest (AESIs). The most common AESIs were proteinuria (60%), hypertension (32.5%), and hemorrhage (32.5%), mainly including gum bleeding and urethrorrhagia. AK109 exhibited an approximately linear PK exposure with dose escalation at 2-12 mg/kg. PD analyses showed rapid target engagement. Among the 40 patients, 4 achieved partial response and 21 achieved stable disease with an objective response rate of 10% and a disease control rate of 62.5%. Based on the safety profile, the PK/PD profile, and preliminary antitumor activities, 12 mg/kg Q2W and 15 mg/kg Q3W were selected as RP2D.

CONCLUSIONS

AK109 showed manageable safety profile and promising antitumor activity, supporting further clinical development in a large population.

UCHL1 Impairs Periodontal Ligament Stem Cell Osteogenesis in Periodontitis

Periodontitis comprises a series of inflammatory responses resulting in alveolar bone loss. The suppression of osteogenesis of periodontal ligament stem cells (PDLSCs) by inflammation is responsible for impaired alveolar bone regeneration, which remains an ongoing challenge for periodontitis therapy. Ubiquitin C-terminal hydrolase L1 (UCHL1) belongs to the family of deubiquitinating enzymes, which was found to play roles in inflammation previously. In this study, the upregulation of UCHL1 was identified in inflamed PDLSCs isolated from periodontitis patients and in healthy PDLSCs treated with tumor necrosis factor-α or interleukin-1β, and the higher expression level of UCHL1 was accompanied with the impaired osteogenesis of PDLSCs. Then UCHL1 was inhibited in PDLSCs using the lentivirus or inhibitor, and the osteogenesis of PDLSCs suppressed by inflammation was rescued by UCHL1 inhibition. Mechanistically, the negative effect of UCHL1 on the osteogenesis of PDLSCs was attributable to its negative regulation of mitophagy-dependent bone morphogenetic protein 2/Smad signaling pathway in periodontitis-associated inflammation. Furthermore, a ligature-induced murine periodontitis model was established, and the specific inhibitor of UCHL1 was administrated to periodontitis mice. The histological results showed increased active osteoblasts on alveolar bone surface and enhanced alveolar bone regeneration when UCHL1 was inhibited in periodontitis mice. Besides, the therapeutic effects of UCHL1 inhibition on ameliorating periodontitis were verified, as indicated by less bone loss and reduced inflammation. Altogether, our study proved UCHL1 to be a key negative regulator of the osteogenesis of PDLSCs in periodontitis and suggested that UCHL1 inhibition holds promise for alveolar bone regeneration in periodontitis treatment.

A comprehensive SARS-CoV-2-human protein-protein interactome reveals COVID-19 pathobiology and potential host therapeutic targets

Studying viral-host protein-protein interactions can facilitate the discovery of therapies for viral infection. We use high-throughput yeast two-hybrid experiments and mass spectrometry to generate a comprehensive SARS-CoV-2-human protein-protein interactome network consisting of 739 high-confidence binary and co-complex interactions, validating 218 known SARS-CoV-2 host factors and revealing 361 novel ones. Our results show the highest overlap of interaction partners between published datasets and of genes differentially expressed in samples from COVID-19 patients. We identify an interaction between the viral protein ORF3a and the human transcription factor ZNF579, illustrating a direct viral impact on host transcription. We perform network-based screens of >2,900 FDA-approved or investigational drugs and identify 23 with significant network proximity to SARS-CoV-2 host factors. One of these drugs, carvedilol, shows clinical benefits for COVID-19 patients in an electronic health records analysis and antiviral properties in a human lung cell line infected with SARS-CoV-2. Our study demonstrates the value of network systems biology to understand human-virus interactions and provides hits for further research on COVID-19 therapeutics.

Interpretable deep learning translation of GWAS and multi-omics findings to identify pathobiology and drug repurposing in Alzheimer's disease

Translating human genetic findings (genome-wide association studies [GWAS]) to pathobiology and therapeutic discovery remains a major challenge for Alzheimer's disease (AD). We present a network topology-based deep learning framework to identify disease-associated genes (NETTAG). We leverage non-coding GWAS loci effects on quantitative trait loci, enhancers and CpG islands, promoter regions, open chromatin, and promoter flanking regions under the protein-protein interactome. Via NETTAG, we identified 156 AD-risk genes enriched in druggable targets. Combining network-based prediction and retrospective case-control observations with 10 million individuals, we identified that usage of four drugs (ibuprofen, gemfibrozil, cholecalciferol, and ceftriaxone) is associated with reduced likelihood of AD incidence. Gemfibrozil (an approved lipid regulator) is significantly associated with 43% reduced risk of AD compared with simvastatin using an active-comparator design (95% confidence interval 0.51-0.63, p < 0.0001). In summary, NETTAG offers a deep learning methodology that utilizes GWAS and multi-genomic findings to identify pathobiology and drug repurposing in AD.

Transcriptomics-based network medicine approach identifies metformin as a repurposable drug for atrial fibrillation

Effective drugs for atrial fibrillation (AF) are lacking, resulting in significant morbidity and mortality. This study demonstrates that network proximity analysis of differentially expressed genes from atrial tissue to drug targets can help prioritize repurposed drugs for AF. Using enrichment analysis of drug-gene signatures and functional testing in human inducible pluripotent stem cell (iPSC)-derived atrial-like cardiomyocytes, we identify metformin as a top repurposed drug candidate for AF. Using the active compactor, a new design analysis of large-scale longitudinal electronic health record (EHR) data, we determine that metformin use is significantly associated with a reduced risk of AF (odds ratio = 0.48, 95%, confidence interval [CI] 0.36-0.64, p < 0.001) compared with standard treatments for diabetes. This study utilizes network medicine methodologies to identify repurposed drugs for AF treatment and identifies metformin as a candidate drug.

ImageGCN: Multi-Relational Image Graph Convolutional Networks for Disease Identification With Chest X-Rays

Image representation is a fundamental task in computer vision. However, most of the existing approaches for image representation ignore the relations between images and consider each input image independently. Intuitively, relations between images can help to understand the images and maintain model consistency over related images, leading to better explainability. In this paper, we consider modeling the image-level relations to generate more informative image representations, and propose ImageGCN, an end-to-end graph convolutional network framework for inductive multi-relational image modeling. We apply ImageGCN to chest X-ray images where rich relational information is available for disease identification. Unlike previous image representation models, ImageGCN learns the representation of an image using both its original pixel features and its relationship with other images. Besides learning informative representations for images, ImageGCN can also be used for object detection in a weakly supervised manner. The experimental results on 3 open-source x-ray datasets, ChestX-ray14, CheXpert and MIMIC-CXR demonstrate that ImageGCN can outperform respective baselines in both disease identification and localization tasks and can achieve comparable and often better results than the state-of-the-art methods.

Sepsis subphenotyping based on organ dysfunction trajectory

BACKGROUND

Sepsis is a heterogeneous syndrome, and the identification of clinical subphenotypes is essential. Although organ dysfunction is a defining element of sepsis, subphenotypes of differential trajectory are not well studied. We sought to identify distinct Sequential Organ Failure Assessment (SOFA) score trajectory-based subphenotypes in sepsis.

METHODS

We created 72-h SOFA score trajectories in patients with sepsis from four diverse intensive care unit (ICU) cohorts. We then used dynamic time warping (DTW) to compute heterogeneous SOFA trajectory similarities and hierarchical agglomerative clustering (HAC) to identify trajectory-based subphenotypes. Patient characteristics were compared between subphenotypes and a random forest model was developed to predict subphenotype membership at 6 and 24 h after being admitted to the ICU. The model was tested on three validation cohorts. Sensitivity analyses were performed with alternative clustering methodologies.

RESULTS

A total of 4678, 3665, 12,282, and 4804 unique sepsis patients were included in development and three validation cohorts, respectively. Four subphenotypes were identified in the development cohort: Rapidly Worsening (n = 612, 13.1%), Delayed Worsening (n = 960, 20.5%), Rapidly Improving (n = 1932, 41.3%), and Delayed Improving (n = 1174, 25.1%). Baseline characteristics, including the pattern of organ dysfunction, varied between subphenotypes. Rapidly Worsening was defined by a higher comorbidity burden, acidosis, and visceral organ dysfunction. Rapidly Improving was defined by vasopressor use without acidosis. Outcomes differed across the subphenotypes, Rapidly Worsening had the highest in-hospital mortality (28.3%, P-value < 0.001), despite a lower SOFA (mean: 4.5) at ICU admission compared to Rapidly Improving (mortality:5.5%, mean SOFA: 5.5). An overall prediction accuracy of 0.78 (95% CI, [0.77, 0.8]) was obtained at 6 h after ICU admission, which increased to 0.87 (95% CI, [0.86, 0.88]) at 24 h. Similar subphenotypes were replicated in three validation cohorts. The majority of patients with sepsis have an improving phenotype with a lower mortality risk; however, they make up over 20% of all deaths due to their larger numbers.

CONCLUSIONS

Four novel, clinically-defined, trajectory-based sepsis subphenotypes were identified and validated. Identifying trajectory-based subphenotypes has immediate implications for the powering and predictive enrichment of clinical trials. Understanding the pathophysiology of these differential trajectories may reveal unanticipated therapeutic targets and identify more precise populations and endpoints for clinical trials.

AX-Unet: A Deep Learning Framework for Image Segmentation to Assist Pancreatic Tumor Diagnosis

Image segmentation plays an essential role in medical imaging analysis such as tumor boundary extraction. Recently, deep learning techniques have dramatically improved performance for image segmentation. However, an important factor preventing deep neural networks from going further is the information loss during the information propagation process. In this article, we present AX-Unet, a deep learning framework incorporating a modified atrous spatial pyramid pooling module to learn the location information and to extract multi-level contextual information to reduce information loss during downsampling. We also introduce a special group convolution operation on the feature map at each level to achieve information decoupling between channels. In addition, we propose an explicit boundary-aware loss function to tackle the blurry boundary problem. We evaluate our model on two public Pancreas-CT datasets, NIH Pancreas-CT dataset, and the pancreas part in medical segmentation decathlon (MSD) medical dataset. The experimental results validate that our model can outperform the state-of-the-art methods in pancreas CT image segmentation. By comparing the extracted feature output of our model, we find that the pancreatic region of normal people and patients with pancreatic tumors shows significant differences. This could provide a promising and reliable way to assist physicians for the screening of pancreatic tumors.

International comparisons of laboratory values from the 4CE collaborative to predict COVID-19 mortality

Given the growing number of prediction algorithms developed to predict COVID-19 mortality, we evaluated the transportability of a mortality prediction algorithm using a multi-national network of healthcare systems. We predicted COVID-19 mortality using baseline commonly measured laboratory values and standard demographic and clinical covariates across healthcare systems, countries, and continents. Specifically, we trained a Cox regression model with nine measured laboratory test values, standard demographics at admission, and comorbidity burden pre-admission. These models were compared at site, country, and continent level. Of the 39,969 hospitalized patients with COVID-19 (68.6% male), 5717 (14.3%) died. In the Cox model, age, albumin, AST, creatine, CRP, and white blood cell count are most predictive of mortality. The baseline covariates are more predictive of mortality during the early days of COVID-19 hospitalization. Models trained at healthcare systems with larger cohort size largely retain good transportability performance when porting to different sites. The combination of routine laboratory test values at admission along with basic demographic features can predict mortality in patients hospitalized with COVID-19. Importantly, this potentially deployable model differs from prior work by demonstrating not only consistent performance but also reliable transportability across healthcare systems in the US and Europe, highlighting the generalizability of this model and the overall approach.

A comprehensive SARS-CoV-2-human protein-protein interactome network identifies pathobiology and host-targeting therapies for COVID-19

Physical interactions between viral and host proteins are responsible for almost all aspects of the viral life cycle and the host's immune response. Studying viral-host protein-protein interactions is thus crucial for identifying strategies for treatment and prevention of viral infection. Here, we use high-throughput yeast two-hybrid and affinity purification followed by mass spectrometry to generate a comprehensive SARS-CoV-2-human protein-protein interactome network consisting of both binary and co-complex interactions. We report a total of 739 high-confidence interactions, showing the highest overlap of interaction partners among published datasets as well as the highest overlap with genes differentially expressed in samples (such as upper airway and bronchial epithelial cells) from patients with SARS-CoV-2 infection. Showcasing the utility of our network, we describe a novel interaction between the viral accessory protein ORF3a and the host zinc finger transcription factor ZNF579 to illustrate a SARS-CoV-2 factor mediating a direct impact on host transcription. Leveraging our interactome, we performed network-based drug screens for over 2,900 FDA-approved/investigational drugs and obtained a curated list of 23 drugs that had significant network proximities to SARS-CoV-2 host factors, one of which, carvedilol, showed promising antiviral properties. We performed electronic health record-based validation using two independent large-scale, longitudinal COVID-19 patient databases and found that carvedilol usage was associated with a significantly lowered probability (17%-20%, P < 0.001) of obtaining a SARS-CoV-2 positive test after adjusting various confounding factors. Carvedilol additionally showed anti-viral activity against SARS-CoV-2 in a human lung epithelial cell line [half maximal effective concentration (EC ₅₀ ) value of 4.1 µM], suggesting a mechanism for its beneficial effect in COVID-19. Our study demonstrates the value of large-scale network systems biology approaches for extracting biological insight from complex biological processes.

Site-specific therapy in cancers of unknown primary site: a systematic review and meta-analysis

Background

Cancer of unknown primary site (CUP) is a term applied to characterize pathologically confirmed metastatic cancer with unknown primary tumor origin. It remains uncertain whether patients with CUP benefit from site-specific therapy guided by molecular profiling.

Patients and methods

A systematic search in PubMed, Web of Science, Embase, Cochrane Library, and ClinicalTrials.gov, and of conference abstracts from January 1976 to January 2021 was performed to identify studies investigating the efficacy of site-specific therapy on patients with CUP. The quality of included studies was evaluated using the Cochrane risk of bias tool and Newcastle–Ottawa scale. Eligible studies were weighted and pooled for meta-analysis. Hazard ratios (HRs) for overall survival (OS) and progression-free survival (PFS) were assessed to compare the efficacy of site-specific therapy with empiric therapy in patients with CUP. In addition, subgroup analyses were conducted.

Results

Five studies comprising 1114 patients were identified, of which 454 patients received site-specific therapy, and 660 patients received empiric therapy. Our meta-analysis revealed that site-specific therapy was not significantly associated with improved PFS [HR 0.93, 95% confidence interval (CI) 0.74-1.17, P = 0.534] and OS (HR 0.75, 95% CI 0.55-1.03, P = 0.069), compared with empiric therapy. However, during subgroup analysis significantly improved OS was associated with site-specific therapy in the high-accuracy predictive assay subgroup (HR 0.46, 95% CI 0.26-0.81, P = 0.008) compared with the low accuracy predictive assay subgroup (HR 0.93, 95% CI 0.75-1.15, P = 0.509). Furthermore, compared with patients with less responsive tumor types, more survival benefit from site-specific therapy was found in patients with more responsive tumors (HR 0.67, 95% CI 0.46-0.97, P = 0.037).

Conclusions

Our results suggest that site-specific therapy is not significantly associated with improved survival outcomes; however, it might benefit patients with CUP with responsive tumor types.

•

Studies evaluating the role of site-specific therapy guided by molecular profiling in CUP provided contradictory results.

•

Site-specific therapy is not significantly associated with improved survival outcomes in the overall CUP population.

•

Molecularly defined site-specific therapy may improve OS only when high-accuracy assays assign CUP to responsive tumor types.

Unique molecular signatures sustained in circulating monocytes and regulatory T cells in Convalescent COVID-19 patients

Over two years into the COVID-19 pandemic, the human immune response to SARS-CoV-2 during the active disease phase has been extensively studied. However, the long-term impact after recovery, which is critical to advance our understanding SARS-CoV-2 and COVID-19-associated long-term complications, remains largely unknown. Herein, we characterized multi-omic single-cell profiles of circulating immune cells in the peripheral blood of 100 patients, including covenlesent COVID-19 and sero-negative controls. The reduced frequencies of both short-lived monocytes and long-lived regulatory T (Treg) cells are significantly associated with the patients recovered from severe COVID-19. Consistently, sc-RNA seq analysis reveals seven heterogeneous clusters of monocytes (M0-M6) and ten Treg clusters (T0-T9) featuring distinct molecular signatures and associated with COVID-19 severity. Asymptomatic patients contain the most abundant clusters of monocyte and Treg expressing high CD74 or IFN-responsive genes. In contrast, the patients recovered from a severe disease have shown two dominant inflammatory monocyte clusters with S100 family genes: S100A8 & A9 with high HLA-I whereas S100A4 & A6 with high HLA-II genes, a specific non-classical monocyte cluster with distinct IFITM family genes, and a unique TGF-β high Treg Cluster. The outpatients and seronegative controls share most of the monocyte and Treg clusters patterns with high expression of HLA genes. Surprisingly, while presumably short-ived monocytes appear to have sustained alterations over 4 months, the decreased frequencies of long-lived Tregs (high HLA-DRA and S100A6) in the outpatients restore over the tested convalescent time (>= 4 months). Collectively, our study identifies sustained and dynamically altered monocytes and Treg clusters with distinct molecular signatures after recovery, associated with COVID-19 severity.

MedGCN: Medication recommendation and lab test imputation via graph convolutional networks

Laboratory testing and medication prescription are two of the most important routines in daily clinical practice. Developing an artificial intelligence system that can automatically make lab test imputations and medication recommendations can save costs on potentially redundant lab tests and inform physicians of a more effective prescription. We present an intelligent medical system (named MedGCN) that can automatically recommend the patients' medications based on their incomplete lab tests, and can even accurately estimate the lab values that have not been taken. In our system, we integrate the complex relations between multiple types of medical entities with their inherent features in a heterogeneous graph. Then we model the graph to learn a distributed representation for each entity in the graph based on graph convolutional networks (GCN). By the propagation of graph convolutional networks, the entity representations can incorporate multiple types of medical information that can benefit multiple medical tasks. Moreover, we introduce a cross regularization strategy to reduce overfitting for multi-task training by the interaction between the multiple tasks. In this study, we construct a graph to associate 4 types of medical entities, i.e., patients, encounters, lab tests, and medications, and applied a graph neural network to learn node embeddings for medication recommendation and lab test imputation. we validate our MedGCN model on two real-world datasets: NMEDW and MIMIC-III. The experimental results on both datasets demonstrate that our model can outperform the state-of-the-art in both tasks. We believe that our innovative system can provide a promising and reliable way to assist physicians to make medication prescriptions and to save costs on potentially redundant lab tests.

Circulating ACE2-expressing extracellular vesicles block broad strains of SARS-CoV-2

The severe acute respiratory syndrome coronavirus 2 (SARS-CoV-2) has caused the pandemic of the coronavirus induced disease 2019 (COVID-19) with evolving variants of concern. It remains urgent to identify novel approaches against broad strains of SARS-CoV-2, which infect host cells via the entry receptor angiotensin-converting enzyme 2 (ACE2). Herein, we report an increase in circulating extracellular vesicles (EVs) that express ACE2 (evACE2) in plasma of COVID-19 patients, which levels are associated with severe pathogenesis. Importantly, evACE2 isolated from human plasma or cells neutralizes SARS-CoV-2 infection by competing with cellular ACE2. Compared to vesicle-free recombinant human ACE2 (rhACE2), evACE2 shows a 135-fold higher potency in blocking the binding of the viral spike protein RBD, and a 60- to 80-fold higher efficacy in preventing infections by both pseudotyped and authentic SARS-CoV-2. Consistently, evACE2 protects the hACE2 transgenic mice from SARS-CoV-2-induced lung injury and mortality. Furthermore, evACE2 inhibits the infection of SARS-CoV-2 variants (α, β, and δ) with equal or higher potency than for the wildtype strain, supporting a broad-spectrum antiviral mechanism of evACE2 for therapeutic development to block the infection of existing and future coronaviruses that use the ACE2 receptor.

[Prediction of 6-year risk of activities of daily living disability in elderly aged 65 years and older in China]

Objective: To construct an easy-to-use risk prediction tool for 6-year risk of activities of daily living(ADL) disability among Chinese elderly aged 65 and above. Methods: A total of 34 349 elderly aged 65 and above were recruited from the Chinese Longitudinal Healthy Longevity Survey. Demographic characteristics, lifestyle and chronic diseases of the elderly were collected through face-to-face interviews. The functional status of the elderly was evaluated by the instrumental activities of daily living(IADL) scale. The mental health status of the elderly was evaluated by the Mini-Mental State Examination. The height, weight, blood pressure and other information of the subjects were obtained through physical examination and body mass index(BMI) was calculated. The ADL status was evaluated by Katz Scale at baseline and follow-up surveys. Taking ADL status as the dependent variable and the key predictors were selected from Lasso regression as the independent variables, a Cox proportional risk regression model was constructed and visualized by the nomogram tool. Area under the receiver operating characteristic curve(AUC) and calibration curve were used to evaluate the discrimination and calibration of the model. A total of 200 bootstrap resamples were used for internal validation of the model. Sensitivity analysis was used to evaluate the robustness of the model. Results: The M(Q₁, Q₃) of subjects' age as 86(75, 94) years old, of which 9 774(46.0%) were males. A total of 112 606 person-years were followed up, 4 578 cases of ADL disability occurred and the incidence density was 40.7/1 000 person-years. Cox proportional risk regression model analysis showed that older age, higher BMI, female, hypertension and history of cerebrovascular disease were associated with higher risk of ADL disability [HR(95%CI) were 1.06(1.05-1.06), 1.05(1.04-1.06), 1.17(1.10-1.25),1.07(1.01-1.13) and 1.41(1.23-1.62), respectively.]; Ethnic minorities, walking 1 km continuously, taking public transportation alone and doing housework almost every day were associated with lower risk of ADL disability [HR(95%CI): 0.71(0.62-0.80), 0.72(0.65-0.80), 0.74(0.68-0.82) and 0.69(0.64-0.74), respectively]. The AUC value of the model was 0.853, and the calibration curve showed that the predicted probability was highly consistent with the observed probability. After excluding non-intervening factors(age, sex and ethnicity), the AUC value of the model for predicting the risk of ADL disability was 0.779. The AUC values of 65-74 years old and 75 years old and above were 0.634 and 0.765, respectively. The AUC values of the model based on walking 1 km continuous and taking public transport alone in IADL and the model based on comprehensive score of IADL were 0.853 and 0.851, respectively. Conclusion: The risk prediction model of ADL disability established in this study has good performance and robustness.

[Association between sleep duration and activity of daily living in the elderly aged 65 years and older in China]

Objective: To investigate the association between sleep duration and activity of daily living (ADL) in the elderly aged 65 years and older in China. Methods: A total of 11 247 subjects aged 65 and above were included in the Chinese Elderly Health Factors Tracking Survey from March 29, 2005 to April 8, 2019. Self-made questionnaire was used to collect the data of population sociological characteristics, health status and disease status. ADL status was assessed by basic activities of daily living. The association between sleep duration and ADL impairment was assessed by Cox proportional risk regression model. The dose-response relationship between sleep duration and ADL impairment was analyzed using restricted cubic spline function. Results: The age of the subjects was (79±10) years, including 5 793(51.5%) females. The incidence of ADL impairment was 33.3% (3 747/11 247). Subjects were divided into short, medium, and long sleep groups according to sleep duration of fewer than seven hours, seven to eight hours, or more than eight hours. The number of short, medium and long sleepers was 2 974 (26.4%), 4 922 (43.8%) and 3 351(29.8%), respectively. The intermediate sleep group had the lowest incidence of impaired ADL (4.98/100 person-years). Cox proportional risk regression model analysis showed that: taking the intermediate sleep group as reference, after adjustment of gender, age, marital status, educational level, place of residence, living with family, smoking, drinking, exercise, frequency of fruit consumption, vegetable intake frequency, sleep quality, factors such as hypertension, diabetes, heart disease and cerebrovascular disease, the long sleep time increased the risk of impaired ADL [HR (95%CI): 1.148 (1.062-1.241)]. Subgroup analysis showed a weak positive multiplicative interaction between sleep duration and age [HR (95%CI): 1.004 (1.000-1.009)], but no multiplicative interaction between sleep duration and sex [HR(95%CI): 0.948 (0.870-1.034)]. Longer sleep duration increased the risk of ADL impairment in women [HR (95%CI): 1.195 (1.074-1.329)], but not in men [HR (95%CI): 1.084 (0.966-1.217)]. Longer sleep duration increased the risk of ADL impairment in people aged 80 years and older [HR (95%CI): 1.185 (1.076-1.305)], but not in people younger than 80 years [HR (95%CI): 1.020 (0.890-1.169)]. There was a non-linear dose-response relationship between sleep duration and ADL damage (P=0.007), and the risk of ADL damage was lowest when sleep duration was 7.5 h. Conclusion: Sleep duration was positively correlated with the risk of ADL impairment in the elderly in a nonlinear dose-response relationship.

Deep learning for cancer type classification and driver gene identification

Background

Genetic information is becoming more readily available and is increasingly being used to predict patient cancer types as well as their subtypes. Most classification methods thus far utilize somatic mutations as independent features for classification and are limited by study power. We aim to develop a novel method to effectively explore the landscape of genetic variants, including germline variants, and small insertions and deletions for cancer type prediction.

Results

We proposed DeepCues, a deep learning model that utilizes convolutional neural networks to unbiasedly derive features from raw cancer DNA sequencing data for disease classification and relevant gene discovery. Using raw whole-exome sequencing as features, germline variants and somatic mutations, including insertions and deletions, were interactively amalgamated for feature generation and cancer prediction. We applied DeepCues to a dataset from TCGA to classify seven different types of major cancers and obtained an overall accuracy of 77.6%. We compared DeepCues to conventional methods and demonstrated a significant overall improvement (p < 0.001). Strikingly, using DeepCues, the top 20 breast cancer relevant genes we have identified, had a 40% overlap with the top 20 known breast cancer driver genes.

Conclusion

Our results support DeepCues as a novel method to improve the representational resolution of DNA sequencings and its power in deriving features from raw sequences for cancer type prediction, as well as discovering new cancer relevant genes.

Supplementary Information

The online version contains supplementary material available at 10.1186/s12859-021-04400-4.

[Sensitivity and specificity of nucleic acid testing in close contacts of COVID-19 cases in Guangzhou]

Objective: To analyze the sensitivity and specificity of SARS-CoV-2 nucleic acid testing in 20 348 close contacts of COVID-19 cases in different prevention and control stages in Guangzhou and to provide scientific evidence for optimizing epidemic response strategies. Methods: A total of 20 348 close contacts of COVID-19 cases in Guangzhou were traced between February 21 and September 22,2020. All the close contacts were tested for the nucleic acid of SARS-CoV-2. The sensitivity and specificity of nucleic acid testing and diagnosis in the different prevention and control stages were compared. Results: In 20 348 close contacts, 12 462 were males (61.24%), the median (P₂₅,P₇₅) of age of them was 31.0 years (23.0,43.0), the median number (P₂₅,P₇₅) of nucleic acid testing for them was 2.0 (1.0,3.0), and the median (P₂₅,P₇₅) of their quarantine days was 12.0 (8.0,13.0) days, respectively. A total of 256 COVID-19 cases were confirmed in the close contacts after seven nucleic acid tests. In the 1^st, 2^nd, 3^rd and 7^th nucleic acid testing, the sensitivity and specificity were 69.14% and 99.99% (177 cases confirmed), 89.84% and 99.99% (230 cases confirmed), 97.27% and 99.99% (249 cases confirmed), and 100.00% and 99.98%, respectively. In the three stages of COVID-19 prevention and control in China: domestic case stage, imported case stage, and imported case associated local epidemic stage, the sensitivity of the 1^st nucleic acid testing was 70.68%, 68.00% and 67.35%, and the specificity was 99.98%, 100.00% and 100.00%, respectively. Conclusions: The sensitivity of nucleic acid testing in the close contacts at the different stages were consistent with slight decrease, which might be related to the increased proportion of asymptomatic infections in the late stage of epidemic prevention and control with COVID-19 in Guangzhou. It is suggested to give three nucleic acid tests to improve the sensitivity and reduce false negative risk.

International Analysis of Electronic Health Records of Children and Youth Hospitalized With COVID-19 Infection in 6 Countries

IMPORTANCE

Additional sources of pediatric epidemiological and clinical data are needed to efficiently study COVID-19 in children and youth and inform infection prevention and clinical treatment of pediatric patients.

OBJECTIVE

To describe international hospitalization trends and key epidemiological and clinical features of children and youth with COVID-19.

DESIGN, SETTING, AND PARTICIPANTS

This retrospective cohort study included pediatric patients hospitalized between February 2 and October 10, 2020. Patient-level electronic health record (EHR) data were collected across 27 hospitals in France, Germany, Spain, Singapore, the UK, and the US. Patients younger than 21 years who tested positive for COVID-19 and were hospitalized at an institution participating in the Consortium for Clinical Characterization of COVID-19 by EHR were included in the study.

MAIN OUTCOMES AND MEASURES

Patient characteristics, clinical features, and medication use.

RESULTS

There were 347 males (52%; 95% CI, 48.5-55.3) and 324 females (48%; 95% CI, 44.4-51.3) in this study's cohort. There was a bimodal age distribution, with the greatest proportion of patients in the 0- to 2-year (199 patients [30%]) and 12- to 17-year (170 patients [25%]) age range. Trends in hospitalizations for 671 children and youth found discrete surges with variable timing across 6 countries. Data from this cohort mirrored national-level pediatric hospitalization trends for most countries with available data, with peaks in hospitalizations during the initial spring surge occurring within 23 days in the national-level and 4CE data. A total of 27 364 laboratory values for 16 laboratory tests were analyzed, with mean values indicating elevations in markers of inflammation (C-reactive protein, 83 mg/L; 95% CI, 53-112 mg/L; ferritin, 417 ng/mL; 95% CI, 228-607 ng/mL; and procalcitonin, 1.45 ng/mL; 95% CI, 0.13-2.77 ng/mL). Abnormalities in coagulation were also evident (D-dimer, 0.78 ug/mL; 95% CI, 0.35-1.21 ug/mL; and fibrinogen, 477 mg/dL; 95% CI, 385-569 mg/dL). Cardiac troponin, when checked (n = 59), was elevated (0.032 ng/mL; 95% CI, 0.000-0.080 ng/mL). Common complications included cardiac arrhythmias (15.0%; 95% CI, 8.1%-21.7%), viral pneumonia (13.3%; 95% CI, 6.5%-20.1%), and respiratory failure (10.5%; 95% CI, 5.8%-15.3%). Few children were treated with COVID-19-directed medications.

CONCLUSIONS AND RELEVANCE

This study of EHRs of children and youth hospitalized for COVID-19 in 6 countries demonstrated variability in hospitalization trends across countries and identified common complications and laboratory abnormalities in children and youth with COVID-19 infection. Large-scale informatics-based approaches to integrate and analyze data across health care systems complement methods of disease surveillance and advance understanding of epidemiological and clinical features associated with COVID-19 in children and youth.

[Trend analysis and prediction of colorectal cancer morbidity and mortality of residents in urban areas of Guangzhou from 1972 to 2015]

Objective: To analyze the trend of mortality and incidence of colorectal cancer among urban residents in Guangzhou from 1972 to 2015 and to predict the mortality of colorectal cancer from 2016 to 2025. Methods: The mortality data of colorectal cancer among urban residents in Guangzhou were collected from the death registration of malignant tumors of Guangzhou Health Statistics Bureau (1972-1979), Guangzhou Health Statistics (1980-2001), Guangzhou Cancer Registration Annual Report (2002-2009) and China Cancer Registration Annual Report (2010-2015). The incidence of colorectal cancer was collected from Guangzhou Cancer Registration Annual Report (2002-2009) and China Cancer Registration Annual Report (2010-2015). The incidence and mortality data of colorectal cancer coded as C18-C21 in 10th Edition of International Classification of Diseases (ICD-10) were obtained from the above data, and the demographic data were from the Guangzhou Municipal Bureau of Statistics. Joinpoint model was used to calculate the annual change percentage (APC) and average annual change percentage (AAPC) of colorectal cancer mortality and incidence among urban residents in Guangzhou from 1972 to 2015 and from 2002 to 2015. ARIMA model was used to predict colorectal cancer mortality from 2016 to 2025. Results: There were 19 309 colorectal cancer deaths among urban residents in Guangzhou from 1972 to 2015. The crude mortality rate of colorectal cancer increased from 4.33/100 000 to 24.89/100 000 (AAPC=4.2%, P<0.001). A total of 24 033 new cases of colorectal cancer were reported in Guangzhou from 2002 to 2015. The crude incidence rate of colorectal cancer increased from 22.95/100 000 to 52.81/100 000 (AAPC=6.6%, P<0.001). The mortality rate of colorectal cancer among urban residents of Guangzhou would continuously increase from 2016 to 2025 and reach 29.53/100 000 in 2025. Conclusion: The mortality rate of colorectal cancer among urban residents of Guangzhou from 1972 to 2015 and the incidence rate of colorectal cancer from 2002 to 2015 both show an upward trend. The mortality rate will increase from 2016 to 2025.

Traditional Chinese medicine clinical records classification with BERT and domain specific corpora

Traditional Chinese Medicine (TCM) has been developed for several thousand years and plays a significant role in health care for Chinese people. This paper studies the problem of classifying TCM clinical records into 5 main disease categories in TCM. We explored a number of state-of-the-art deep learning models and found that the recent Bidirectional Encoder Representations from Transformers can achieve better results than other deep learning models and other state-of-the-art methods. We further utilized an unlabeled clinical corpus to fine-tune the BERT language model before training the text classifier. The method only uses Chinese characters in clinical text as input without preprocessing or feature engineering. We evaluated deep learning models and traditional text classifiers on a benchmark data set. Our method achieves a state-of-the-art accuracy 89.39% ± 0.35%, Macro F1 score 88.64% ± 0.40% and Micro F1 score 89.39% ± 0.35%. We also visualized attention weights in our method, which can reveal indicative characters in clinical text.

International Comparisons of Harmonized Laboratory Value Trajectories to Predict Severe COVID-19: Leveraging the 4CE Collaborative Across 342 Hospitals and 6 Countries: A Retrospective Cohort Study

Objectives

To perform an international comparison of the trajectory of laboratory values among hospitalized patients with COVID-19 who develop severe disease and identify optimal timing of laboratory value collection to predict severity across hospitals and regions.

Design

Retrospective cohort study.

Setting

The Consortium for Clinical Characterization of COVID-19 by EHR (4CE), an international multi-site data-sharing collaborative of 342 hospitals in the US and in Europe.

Participants

Patients hospitalized with COVID-19, admitted before or after PCR-confirmed result for SARS-CoV-2.

Primary and secondary outcome measures

Patients were categorized as "ever-severe" or "never-severe" using the validated 4CE severity criteria. Eighteen laboratory tests associated with poor COVID-19-related outcomes were evaluated for predictive accuracy by area under the curve (AUC), compared between the severity categories. Subgroup analysis was performed to validate a subset of laboratory values as predictive of severity against a published algorithm. A subset of laboratory values (CRP, albumin, LDH, neutrophil count, D-dimer, and procalcitonin) was compared between North American and European sites for severity prediction.

Results

Of 36,447 patients with COVID-19, 19,953 (43.7%) were categorized as ever-severe. Most patients (78.7%) were 50 years of age or older and male (60.5%). Longitudinal trajectories of CRP, albumin, LDH, neutrophil count, D-dimer, and procalcitonin showed association with disease severity. Significant differences of laboratory values at admission were found between the two groups. With the exception of D-dimer, predictive discrimination of laboratory values did not improve after admission. Sub-group analysis using age, D-dimer, CRP, and lymphocyte count as predictive of severity at admission showed similar discrimination to a published algorithm (AUC=0.88 and 0.91, respectively). Both models deteriorated in predictive accuracy as the disease progressed. On average, no difference in severity prediction was found between North American and European sites.

Conclusions

Laboratory test values at admission can be used to predict severity in patients with COVID-19. Prediction models show consistency across international sites highlighting the potential generalizability of these models.

Surgical resection of clinically benign tumours in the maxillomandibular deep lobe of the parotid gland via sternocleidomastoid muscle-parotid space approach

This article reports the surgical resection of clinically benign tumours in the maxillomandibular deep lobe of the parotid gland via sternocleidomastoid muscle-parotid space (SPS) approach. The use of maxillary-mandibular planes to subdivide the deep lobe of the parotid gland in order to establish the tumour location and accessibility is introduced. This approach, which does not raise a skin flap, may preserve the superficial lobe. Ten patients with clinically benign tumours in the maxillomandibular deep lobe of the parotid gland were treated via the SPS approach. The patients were followed up for 3-5 years and the surgical outcomes were analysed. All tumours were completely enucleated via the SPS approach with an optimal aesthetic outcome. No permanent facial weakness or tumour recurrence was identified during the 3-5 years of follow-up. The SPS approach to surgical resection is an ideal option for clinically benign tumours in the maxillomandibular deep lobe of the parotid gland and demonstrates good results.

[Effects of oxygen saturation on all-cause mortality among the elderly over 65 years old in 9 longevity areas of China]

Objective: To investigate the association between oxygen saturation (SpO₂) and risk of 3-year all-cause mortality among Chinese older adults aged 65 or over. Methods: The participants were enrolled from Healthy Aging and Biomarkers Cohort Study in year of 2012 to 2014 in 9 longevity areas in China. In this prospective cohort study, 2 287 participants aged 65 or over were enrolled. Data on SpO₂ and body measurements were collected at baseline in 2012, and data on survival outcome and time of mortality were collected at the follow-up in 2014. Participants were divided into two groups according to whether SpO₂ was abnormal (SpO₂<94% was defined as abnormal). Results: The 2 287 participants were (86.5±12.2) years old, 1 006 were males (44.0%), and 315 (13.8%) were abnormal in SpO₂. During follow-up in 2014, 452 were died, 1 434 were survived, and 401 were lost to follow-up. The all-cause mortality rate was 19.8%, and the follow-up rate was 82.5%. The mortality rate of SpO₂ in normal group was 21.1%, and that of abnormal group was 41.6% (P<0.001). After adjusting for confounding factors, compared to participants with normal SpO₂, participants with abnormal SpO₂ had increased risk of all-cause mortality with HR (95%CI) of 1.62 (1.31-2.02); HR (95 % CI) was 1.49 (0.98-2.26) for males and 1.71 (1.30-2.26) for females in abnormal SpO₂ group, respectively; HR (95%CI) was 2.70 (0.98-7.44) for aged 65-79 years old, 1.22 (0.63-2.38) for aged 80-89 years old, and 1.72 (1.35-2.19) for aged over 90 years old in abnormal SpO₂ group, respectively. Conclusion: Abnormal SpO₂ was responsible for increased risk of 3-year all-cause mortality among Chinese elderly adults.

[The relationship between resting heart rate and all-cause mortality among the Chinese oldest-old aged more than 80: a prospective cohort study]

Objective: To explore the association between resting heart rate(RHR) and all-cause mortality among the Chinese oldest-old aged more than 80. Methods: Using a total of seven surveys or follow-ups data (1998, 2000, 2002, 2005, 2008, 2011 and 2014) from the Chinese Longitudinal Healthy Longevity Survey (CLHLS). A total of 17 886 elderly over 80 years old were selected as subjects, their resting heart rate were measured though baseline survey and the survival outcome and death time of the subjects were followed up. The subjects were divided into 6 groups according to their resting heart rate. Cox regression model was used to estimate the effect of resting heart rate on mortality risk. The interaction of age, gender and resting heart rate was also analyzed by likelihood ratio test. Results: The age of subjects M (P₂₅, P₇₅) was 92 (86, 100) years old, including 10 531 females (58.9%) and there were 13 598 participants died, the mortality rate was 195.5 per 1 000 person-years. Multivariate Cox regression analysis showed that compared to the control group (60-69 pbm/min), the hazard ratio of the elderly are 1.06 (95%CI: 1.02, 1.11), 1.09 (95%CI: 1.04, 1.15), 1.23 (95%CI: 1.14, 1.34), 1.25 (95%CI: 1.08, 1.44) in the group of RHR between 70-79, 80-89, 90-99 and ≥100 pbm/min and P values are all less than 0.05. Likelihood ratio test showed that RHR and age had an interaction effect. (P for interaction=0.011). Conclusion: The risk of all-cause death increased with the increase of resting heart rate and this relationship was stronger between the 80-89 years old people.

Data-driven discovery of probable Alzheimer's disease and related dementia subphenotypes using electronic health records

Introduction

We sought to assess longitudinal electronic health records (EHRs) using machine learning (ML) methods to computationally derive probable Alzheimer's Disease (AD) and related dementia subphenotypes.

Methods

A retrospective analysis of EHR data from a cohort of 7587 patients seen at a large, multi‐specialty urban academic medical center in New York was conducted. Subphenotypes were derived using hierarchical clustering from 792 probable AD patients (cases) who had received at least one diagnosis of AD using their clinical data. The other 6795 patients, labeled as controls, were matched on age and gender with the cases and randomly selected in the ratio of 9:1. Prediction models with multiple ML algorithms were trained on this cohort using 5‐fold cross‐validation. XGBoost was used to rank the variable importance.

Results

Four subphenotypes were computationally derived. Subphenotype A (n = 273; 28.2%) had more patients with cardiovascular diseases; subphenotype B (n = 221; 27.9%) had more patients with mental health illnesses, such as depression and anxiety; patients in subphenotype C (n = 183; 23.1%) were overall older (mean (SD) age, 79.5 (5.4) years) and had the most comorbidities including diabetes, cardiovascular diseases, and mental health disorders; and subphenotype D (n = 115; 14.5%) included patients who took anti‐dementia drugs and had sensory problems, such as deafness and hearing impairment.

The 0‐year prediction model for AD risk achieved an area under the receiver operating curve (AUC) of 0.764 (SD: 0.02); the 6‐month model, 0.751 (SD: 0.02); the 1‐year model, 0.752 (SD: 0.02); the 2‐year model, 0.749 (SD: 0.03); and the 3‐year model, 0.735 (SD: 0.03), respectively. Based on variable importance, the top‐ranked comorbidities included depression, stroke/transient ischemic attack, hypertension, anxiety, mobility impairments, and atrial fibrillation. The top‐ranked medications included anti‐dementia drugs, antipsychotics, antiepileptics, and antidepressants.

Conclusions

Four subphenotypes were computationally derived that correlated with cardiovascular diseases and mental health illnesses. ML algorithms based on patient demographics, diagnosis, and treatment demonstrated promising results in predicting the risk of developing AD at different time points across an individual's lifespan.

Immunogenicity and protective ability of RpoE against Streptococcus suis serotype 2

AIMS

RpoE is quite immunogenic and can be used as a candidate vaccine for Streptococcus suis infection via immunoproteomics as reported in our previous studies. In this study, we aimed to verify the immunogenicity of recombinant RpoE and its protective effect against of S. suis.

METHODS AND RESULTS

The RpoE protein was successfully expressed in Escherichia coli, and the purified recombinant protein was mixed with ISA206 to prepare an S. suis subunit vaccine. Mice were immunized with the RpoE subunit vaccine and then infected with the virulent S. suis strain ZY05719. Subunit vaccine-immunized mice achieved 50% protection, less pathological damage and less bacterial distribution in each organ compared with the control mice. Furthermore, in vitro culture, showed that mouse antisera significantly (P ＜ 0·001) inhibited the growth of S. suis, and qRT-PCR results showed that RpoE successfully induced the up-regulation of IL-6 and TNF-α cytokines.

CONCLUSIONS

RpoE mice were vaccinated to obtain immune protection, which may be candidates for S. suis subunit vaccine.

SIGNIFICANCE AND IMPACT OF THE STUDY

The results of this study will provide new ideas for the development of safe and effective recombinant subunits vaccines for S. suis.

Does the Brown classification of maxillectomy defects have prognostic prediction for patients with oral cavity squamous cell carcinoma involving the maxilla?

The aim of this study was to investigate the correlation between the maxillectomy defect, T stage, and prognosis of patients with maxillary squamous cell carcinoma (SCC). The Brown classification system was used to appraise the maxillectomy defects due to maxillary SCC. The clinical data of 137 patients with maxillary SCC during the period 2000-2010 were reviewed; 105 patients were followed up. Preoperative T stage and postoperative maxillectomy class were recorded. The relationship between the maxillectomy defect class and T stage of maxillary SCC was analysed. Correlations between the maxillectomy defect class, local recurrence rate, and survival rate were assessed using IBM SPSS Statistics v19.0. The most common maxillectomy defect class was IIb (54.7%, 75/137). The maxillectomy defect class was significantly associated with the T stage (P < 0.001). Both T stage and the maxillectomy defect class were significantly associated with the survival rate of patients with maxillary SCC (both P< 0.001). In conclusion, the class of the maxillectomy defect was found to be associated with the T stage. Both of these were prognostic factors for patients with maxillary SCC. The class of the maxillectomy defect is suitable for clinical application in predicting the prognosis compared with T stage.

Integrating hypertension phenotype and genotype with hybrid non-negative matrix factorization

MOTIVATION

Hypertension is a heterogeneous syndrome in need of improved subtyping using phenotypic and genetic measurements with the goal of identifying subtypes of patients who share similar pathophysiologic mechanisms and may respond more uniformly to targeted treatments. Existing machine learning approaches often face challenges in integrating phenotype and genotype information and presenting to clinicians an interpretable model. We aim to provide informed patient stratification based on phenotype and genotype features.

RESULTS

In this article, we present a hybrid non-negative matrix factorization (HNMF) method to integrate phenotype and genotype information for patient stratification. HNMF simultaneously approximates the phenotypic and genetic feature matrices using different appropriate loss functions, and generates patient subtypes, phenotypic groups and genetic groups. Unlike previous methods, HNMF approximates phenotypic matrix under Frobenius loss, and genetic matrix under Kullback-Leibler (KL) loss. We propose an alternating projected gradient method to solve the approximation problem. Simulation shows HNMF converges fast and accurately to the true factor matrices. On a real-world clinical dataset, we used the patient factor matrix as features and examined the association of these features with indices of cardiac mechanics. We compared HNMF with six different models using phenotype or genotype features alone, with or without NMF, or using joint NMF with only one type of loss We also compared HNMF with 3 recently published methods for integrative clustering analysis, including iClusterBayes, Bayesian joint analysis and JIVE. HNMF significantly outperforms all comparison models. HNMF also reveals intuitive phenotype-genotype interactions that characterize cardiac abnormalities.

AVAILABILITY AND IMPLEMENTATION

Our code is publicly available on github at https://github.com/yuanluo/hnmf.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

[Effects of estimated glomerular filtration rate on all-cause mortality in the elderly aged 65 years and older in 8 longevity areas in China]

Objective: To investigate the association between estimated glomerular filtration rate (eGFR) and all-cause mortality in the elderly aged 65 years and older in longevity areas in China. Methods: Data used in this study were obtained from Healthy Aging and Biomarkers Cohort Study, a sub-cohort of the Chinese Longitudinal Healthy Longevity Survey, 1 802 elderly adults were collected in the study during 2012-2017/2018. In this study, the elderly were classified into 4 groups, moderate-to-severe group [<45 ml·min(-1)·(1.73 m(2))(-1)], mild-to-moderate group [45- ml·min(-1)·(1.73 m(2))(-1)], mild group [60- ml·min(-1)·(1.73 m(2))(-1)] and normal group [≥90 ml·min(-1)·(1.73 m(2))(-1)] according to their eGFR levels. Results: After 6 years of follow-up, 852 participants died, with a mortality rate of 47.3%. Multivariate Cox regression analysis showed that the levels of eGFR were negatively correlated with all-cause mortality risk in the elderly (the HR of elderly was 0.993 and the 95%CI was 0.989-0.997 for every unit of eGFR increased, P=0.001), while compared with the group with normal eGFR, the HRs (95%CI) of the elderly in the moderate-to-severe group, mild-to-moderate group, and mild group were 1.690 (1.224-2.332, P=0.001), 1.312 (0.978-1.758, P=0.070), 1.349 (1.047-1.737, P=0.020) respectively [trend test P<0.001]. Conclusion: The decrease in eGFR was associated with higher mortality risk among the elderly in longevity areas in China.

[Historical perspective of progress and achievement on epidemiology in the past 70 years in China]

Epidemiology is a discipline developed in the practice of preventing diseases and promoting health and is the key of public health and preventive medicine. Since the founding of the People's Republic of China, with the changing of disease pattern in populations, the applications of epidemiology now have expanded from infectious diseases to chronic non-communicable diseases, injuries and health related events. The discipline has made remarkable achievements in the field of disease prevention and control, scientific research and teaching, institution building and academic journals. In this paper we briefly review the history and achievements of epidemiology in China in the past 70 years, and explore the future development of the discipline, which may leave a trace of history for the development of epidemiology in China.

Graph Convolutional Networks for Text Classification

Text classification is an important and classical problem in natural language processing. There have been a number of studies that applied convolutional neural networks (convolution on regular grid, e.g., sequence) to classification. However, only a limited number of studies have explored the more flexible graph convolutional neural networks (convolution on non-grid, e.g., arbitrary graph) for the task. In this work, we propose to use graph convolutional networks for text classification. We build a single text graph for a corpus based on word co-occurrence and document word relations, then learn a Text Graph Convolutional Network (Text GCN) for the corpus. Our Text GCN is initialized with one-hot representation for word and document, it then jointly learns the embeddings for both words and documents, as supervised by the known class labels for documents. Our experimental results on multiple benchmark datasets demonstrate that a vanilla Text GCN without any external word embeddings or knowledge outperforms state-of-the-art methods for text classification. On the other hand, Text GCN also learns predictive word and document embeddings. In addition, experimental results show that the improvement of Text GCN over state-of-the-art comparison methods become more prominent as we lower the percentage of training data, suggesting the robustness of Text GCN to less training data in text classification.

Cancer classification and pathway discovery using non-negative matrix factorization

OBJECTIVES

Extracting genetic information from a full range of sequencing data is important for understanding disease. We propose a novel method to effectively explore the landscape of genetic mutations and aggregate them to predict cancer type.

DESIGN

We applied non-smooth non-negative matrix factorization (nsNMF) and support vector machine (SVM) to utilize the full range of sequencing data, aiming to better aggregate genetic mutations and improve their power to predict disease type. More specifically, we introduce a novel classifier to distinguish cancer types using somatic mutations obtained from whole-exome sequencing data. Mutations were identified from multiple cancers and scored using SIFT, PP2, and CADD, and collapsed at the individual gene level. nsNMF was then applied to reduce dimensionality and obtain coefficient and basis matrices. A feature matrix was derived from the obtained matrices to train a classifier for cancer type classification with the SVM model.

RESULTS

We have demonstrated that the classifier was able to distinguish four cancer types with reasonable accuracy. In five-fold cross-validations using mutation counts as features, the average prediction accuracy was 80% (SEM = 0.1%), significantly outperforming baselines and outperforming models using mutation scores as features.

CONCLUSION

Using the factor matrices derived from the nsNMF, we identified multiple genes and pathways that are significantly associated with each cancer type. This study presents a generic and complete pipeline to study the associations between somatic mutations and cancers. The proposed method can be adapted to other studies for disease status classification and pathway discovery.

[Association of plasma albumin and hypersensitive C-reactive protein with 5-year all-cause mortality among Chinese older adults aged 65 and older from 8 longevity areas in China]

Objective: To investigate the relationship of plasma albumin and hypersensitive C-reactive protein (Hs-CRP) with 5-year all-cause mortality among Chinese older adults aged 65 and older. Method: Data was collected in 8 longevity areas of the Chinese Longitudinal Healthy Longevity Survey (CLHLS) study conducted by Chinese Center for Disease Control and Prevention and Peking University at baseline survey in 2012 and 2014, the participants enrolled in 2012 was followed-up in 2014 and 2017, the participants enrolled in 2014 was followed-up in 2017 only. Finally, 3 118 older adults aged 65 and older with complete information on albumin, Hs-CRP and body mass index (BMI) were included in this study. Plasma samples of older adults were collected for the detection of albumin and Hs-CRP at baseline survey. Survival status and follow-up time was recorded for all participants. All older adults were divided into 4 groups according to the levels of plasma albumin and Hs-CRP, and Cox proportional hazard models were constructed to assess their influence on the risk of all-cause mortality. Results: Among 3 118 older adults included, the prevalence of hypoalbuminemia was 10.1% (316/3 118), and was 22.8% (711/3 118) for elevated Hs-CRP. During 10 132 person-years of follow-up, 1 212 participants died. Participants with hypoalbuminemia had increased risk of all-cause mortality, with an hazard ratio (HR) and 95% confidential interval (CI) of 1.18 (1.01-1.38), compared to participants with normal plasma albuminemia; participants with elevated Hs-CRP had increased risk of all-cause mortality, with an HR (95%CI) of 1.18 (1.04-1.35), compared to participants with normal plasma Hs-CRP. Participants with normal plasma albumin and elevated Hs-CRP, with hypoalbuminemia and normal Hs-CRP, with hypoalbuminemia and elevated Hs-CRP also had increased risk of all-cause mortality when compared to those with normal plasma albumin and normal Hs-CRP, the HR (95%CI) were 1.16 (1.01-1.34), 1.11 (0.91-1.37) and 1.43 (1.11-1.83), respectively. Conclusion: Hypoalbuminemia and elevated Hs-CRP were responsible for increased risk of 5-year all-cause mortality among Chinese older adults from 8 longevity areas.

[The relationship between hazard ratio and median survival time]

The hazard ratio and median survival time are the routine indicators in survival analysis. We briefly introduced the relationship between hazard ratio and median survival time and the role of proportional hazard assumption. We compared 110 pairs of hazard ratio and median survival time ratio in 58 articles and demonstrated the reasons for the difference by examples. The results showed that the hazard ratio estimated by the Cox regression model is unreasonable and not equivalent to median survival time ratio when the proportional hazard assumption is not met. Therefore, before performing the Cox regression model, the proportional hazard assumption should be tested first. If proportional hazard assumption is met, Cox regression model can be used; if proportional hazard assumption is not met, restricted mean survival times is suggested.

Developing a portable natural language processing based phenotyping system

BACKGROUND

This paper presents a portable phenotyping system that is capable of integrating both rule-based and statistical machine learning based approaches.

METHODS

Our system utilizes UMLS to extract clinically relevant features from the unstructured text and then facilitates portability across different institutions and data systems by incorporating OHDSI's OMOP Common Data Model (CDM) to standardize necessary data elements. Our system can also store the key components of rule-based systems (e.g., regular expression matches) in the format of OMOP CDM, thus enabling the reuse, adaptation and extension of many existing rule-based clinical NLP systems. We experimented with our system on the corpus from i2b2's Obesity Challenge as a pilot study.

RESULTS

Our system facilitates portable phenotyping of obesity and its 15 comorbidities based on the unstructured patient discharge summaries, while achieving a performance that often ranked among the top 10 of the challenge participants.

CONCLUSION

Our system of standardization enables a consistent application of numerous rule-based and machine learning based classification techniques downstream across disparate datasets which may originate across different institutions and data systems.