Uncovering Clinical Risk Factors and Predicting Severe COVID-19 Cases Using UK Biobank Data: Machine Learning Approach

Background

COVID-19 is a major public health concern. Given the extent of the pandemic, it is urgent to identify risk factors associated with disease severity. More accurate prediction of those at risk of developing severe infections is of high clinical importance.

Objective

Based on the UK Biobank (UKBB), we aimed to build machine learning models to predict the risk of developing severe or fatal infections, and uncover major risk factors involved.

Methods

We first restricted the analysis to infected individuals (n=7846), then performed analysis at a population level, considering those with no known infection as controls (ncontrols=465,728). Hospitalization was used as a proxy for severity. A total of 97 clinical variables (collected prior to the COVID-19 outbreak) covering demographic variables, comorbidities, blood measurements (eg, hematological/liver/renal function/metabolic parameters), anthropometric measures, and other risk factors (eg, smoking/drinking) were included as predictors. We also constructed a simplified (lite) prediction model using 27 covariates that can be more easily obtained (demographic and comorbidity data). XGboost (gradient-boosted trees) was used for prediction and predictive performance was assessed by cross-validation. Variable importance was quantified by Shapley values (ShapVal), permutation importance (PermImp), and accuracy gain. Shapley dependency and interaction plots were used to evaluate the pattern of relationships between risk factors and outcomes.

Results

A total of 2386 severe and 477 fatal cases were identified. For analyses within infected individuals (n=7846), our prediction model achieved area under the receiving-operating characteristic curve (AUC–ROC) of 0.723 (95% CI 0.711-0.736) and 0.814 (95% CI 0.791-0.838) for severe and fatal infections, respectively. The top 5 contributing factors (sorted by ShapVal) for severity were age, number of drugs taken (cnt_tx), cystatin C (reflecting renal function), waist-to-hip ratio (WHR), and Townsend deprivation index (TDI). For mortality, the top features were age, testosterone, cnt_tx, waist circumference (WC), and red cell distribution width. For analyses involving the whole UKBB population, AUCs for severity and fatality were 0.696 (95% CI 0.684-0.708) and 0.825 (95% CI 0.802-0.848), respectively. The same top 5 risk factors were identified for both outcomes, namely, age, cnt_tx, WC, WHR, and TDI. Apart from the above, age, cystatin C, TDI, and cnt_tx were among the top 10 across all 4 analyses. Other diseases top ranked by ShapVal or PermImp were type 2 diabetes mellitus (T2DM), coronary artery disease, atrial fibrillation, and dementia, among others. For the “lite” models, predictive performances were broadly similar, with estimated AUCs of 0.716, 0.818, 0.696, and 0.830, respectively. The top ranked variables were similar to above, including age, cnt_tx, WC, sex (male), and T2DM.

Conclusions

We identified numerous baseline clinical risk factors for severe/fatal infection by XGboost. For example, age, central obesity, impaired renal function, multiple comorbidities, and cardiometabolic abnormalities may predispose to poorer outcomes. The prediction models may be useful at a population level to identify those susceptible to developing severe/fatal infections, facilitating targeted prevention strategies. A risk-prediction tool is also available online. Further replications in independent cohorts are required to verify our findings.

[Summary and prospect of early warning models and systems for infectious disease outbreaks]

This paper summarizes the basic principles and models of early warning for infectious disease outbreaks, introduces the early warning systems for infectious disease based on different data sources and their applications, and discusses the application potential of big data and their analysing techniques, which have been studied and used in the prevention and control of COVID-19 pandemic, including internet inquiry, social media, mobile positioning, in the early warning of infectious diseases in order to provide reference for the establishment of an intelligent early warning mechanism and platform for infectious diseases based on multi-source big data.

Association of improved overall survival with decreased distant metastasis following asparaginase-based chemotherapy and radiotherapy for intermediate- and high-risk early-stage extranodal nasal-type NK/T-cell lymphoma: a CLCG study

BACKGROUND

This study evaluated the survival benefit of asparaginase (ASP)-based versus non-ASP-based chemotherapy combined with radiotherapy in a real-world cohort of patients with early-stage extranodal nasal-type natural killer/T-cell lymphoma (ENKTCL).

PATIENTS AND METHODS

We identified 376 patients who received combined radiotherapy with either ASP-based (ASP, platinum, and gemcitabine; n = 286) or non-ASP-based (platinum and gemcitabine; n = 90) regimens. The patients were stratified into low-, intermediate-, and high-risk groups using the early stage-adjusted nomogram-revised risk index. Overall survival (OS) and distant metastasis (DM)-free survival (DMFS) between the chemotherapy regimens were compared using inverse probability of treatment weighting (IPTW) and multivariable analyses.

RESULTS

ASP-based (versus non-ASP-based) regimens significantly improved 5-year OS (84.5% versus 73.2%, P = 0.021) and DMFS (84.4% versus 74.5%, P = 0.014) for intermediate- and high-risk patients, but not for low-risk patients in the setting of radiotherapy. Moreover, ASP-based regimens decreased DM, with a 5-year cumulative DM rate of 14.9% for ASP-based regimens compared with 25.1% (P = 0.014) for non-ASP-based regimens. The survival benefit of ASP-based chemotherapy and radiotherapy remained consistent after adjusting the confounding variables using IPTW and multivariate analyses; additional sensitivity analyses confirmed these results.

CONCLUSIONS

The findings provided support for ASP-based chemotherapy and radiotherapy as a first-line treatment strategy for intermediate- and high-risk early-stage ENKTCL.

A framework to decipher the genetic architecture of combinations of complex diseases: Applications in cardiovascular medicine

MOTIVATION

Currently, most genome-wide association studies (GWAS) are studies of a single disease against controls. However, an individual is often affected by more than one condition. For example, coronary artery disease (CAD) is often comorbid with type 2 diabetes (T2DM). Similarly, it is clinically meaningful to study patients with one disease but without a related comorbidity. For example, obese T2DM may have different pathophysiology from non-obese T2DM.

RESULTS

We developed a statistical framework (CombGWAS) to uncover susceptibility variants for comorbid disorders (or a disorder without comorbidity), using GWAS summary statistics only. In essence, we mimicked a case-control GWAS in which the cases are affected with comorbidities or a disease without comorbidity. We extended our methodology to analyze continuous traits with clinically meaningful categories (e.g. lipids), and combination of more than 2 traits.We verified the feasibility and validity of our method by applying it to simulated scenarios and four cardiometabolic (CM) traits. In total, we identified 384 and 587 genomic risk loci respectively for 6 comorbidities and 12 CM disease 'subtypes' without a relevant comorbidity. Genetic correlation analysis revealed that some subtypes may be biologically distinct from others. Further Mendelian randomization analysis showed differential causal effects of different subtypes to relevant complications. For example, we found that obese T2DM is causally related to increased risk of CAD (p=2.62E-11).

AVAILABILITY

The R code is available at: https://github.com/LiangyingYin/CombGWAS.

SUPPLEMENTARY INFORMATION

Supplementary data are available at Bioinformatics online.

Functional study of a N-terminal CPVT mutation RyR2R420Q in patient specific hiPSC-CMs model

Funding Acknowledgements

Type of funding sources: Public grant(s) – National budget only. Main funding source(s): ANR (Agence Nationale de la Rercherche)

Catecholaminergic polymorphic ventricular tachycardia (CPVT) is a lethal genetic arrhythmia that manifests by syncope or sudden death in children and young adults under stress conditions without obvious cardiac structural abnormality. A novel CPVT mutation located in the RyR2 N terminal portion has been identified in a Spanish family (RyR2R420Q). According to the studies of RyR2 function in HEK293 cell line, this mutation presented gain of function at low cytosolic intracellular Ca2+ concentration ([Ca2+]i) and loss of function at high [Ca2+]i. Moreover, KI mice heterozygous for this mutation presented bradycardia and sino-atrial node (SAN) dysfunction. Here we generated induced pluripotent stem cell (hiPSC) from two brothers (one with mutation, the other without mutation as control) of this family and differentiated them into cardiomyocytes (hiPSC-CM). In order to verify that the differentiated cells were well cardiomyocytes, we did immunofluorescence labelling to detect the α-actinin expression and found that around 90% cells were α-actinin positive in both groups of hiPS-CMs. Then the calcium transient was studied by confocal microscopy and the action potential (AP) by micro-electrode technique. The characteristics of spontaneous AP of mutated cells were mostly similar to that of control cells, but more mutated cells presented proarrhythmic behaviors under adrenergic stimulation. hiPSC-CM are immature cardiomyocytes and contract spontaneously. In order to be able to analyze [Ca2+]i transient characteristics, we paced the cells at a constant rate of 1 Hz by field stimulation through two Pt electrodes. Sarcoplasmic reticulum (SR) Ca2+ load was estimated by rapid caffeine (10 mM) application. hiPSC-CMs from the RyR2R420Q carrier presented smaller SR Ca2+ load than those from the control person, whereas their fractional release (the [Ca2+]i transient normalized by the amount of Ca2+ stored in the SR) was higher than that in control group, indicating a gain-of-function mutation.  Even if SR Ca2+ load was smaller in RyR2R420Q cells, they often presented proarrhythmogenic behavior such as Ca2+ waves. The fact was further enhanced during β-adrenergic stimulation, pointing to this model as a valuable tool to study the CPVT disease in human cells.

The potential role of plasma fibroblast growth factor 21 as a diagnostic biomarker for abdominal aortic aneurysm presence and development

AIMS

Fibroblast growth factor 21 (FGF21) has been identified as the master hormonal regulator of energy balance, its elevation is observed in a series of metabolic and cardiovascular diseases. Studies have implicated the role of FGF21 signaling in the pathogenesis of abdominal aortic aneurysm (AAA). We will investigate the association of FGF21 and AAA development.

MATERIALS AND METHODS

In this study, we assayed plasma levels of FGF21 in 82 patients with AAA and 44 control subjects, then analyzed their relationship with clinical, biochemical and histological phenotypes. The expression of β-klotho, an essential co-receptor of FGF21, was assessed with IHC staining and RT-qPCR. Machine learning models incorporate a combination of FGF21 and clinical data were utilized in the prediction of AAA occurrence.

KEY FINDINGS

FGF21 was statistically higher in patients with AAA (781 pg/ml [533, 1213]) than in control subjects (567 pg/ml [324, 939]). After adjustment for age and BMI, we found a positive association of FGF21 levels with AAA diameters, hypertension rate and hsCRP, and a negative correlation between FGF21 levels and HDL-c. Furthermore, the protein levels of β-klotho in abdominal aorta of AAA were found significantly lower than in control group indicating the presence of FGF21 resistance. Combining FGF21 levels with four clinical characteristics significantly improved the stratification of AAA and control groups with an AUC of 0.778.

SIGNIFICANCE

Combining detection of plasma FGF21 and clinical characteristics may be reliable for identifying the presence of AAA. The role of FGF21 as a therapeutic target of AAA warrants further investigation.

Cross-Beam Energy Transfer Saturation by Ion Heating

We measure cross-beam energy transfer (CBET) saturation by ion heating in a gas-jet plasma characterized using Thomson scattering. A wavelength-tunable ultraviolet (UV) probe laser beam interacts with four intense UV pump beams to drive large-amplitude ion-acoustic waves. For the highest-intensity interactions, the power transfer to the probe laser drops, demonstrating ion-acoustic wave saturation. Over this time, the ion temperature is measured to increase by a factor of 7 during the 500-ps interaction. Particle-in-cell simulations show ion trapping and a subsequent ion heating consistent with measurements. Linear kinetic CBET models are found to agree well with the observed energy transfer when the measured plasma conditions are used.

Improving probiotic spore yield using rice straw hydrolysate

Spore-forming Bacillus sp. has been extensively studied for their probiotic properties. In this study, an acid-treated rice straw hydrolysate was used as carbon source to produce the spores of Bacillus coagulans. The results showed that this hydrolysate significantly improved the spore yield compared with other carbon sources such as glucose. Three significant medium components including rice straw hydrolysate, MnSO₄ and yeast extract were screened by Plackett-Burman design. These significant variables were further optimized by response surface methodology (RSM). The optimal values of the medium components were rice straw hydolysate of 27% (v/v), MnSO₄ of 0·78 g l^-1 and yeast extract of 1·2 g l^-1 . The optimized medium and RSM model for spore production were validated in a 5 l bioreactor. Overall, this sporulation medium containing acid-treated rice straw hydrolysate has a potential to be used in the production of B. coagulans spores.

MTR D919G variant is associated with prostate adenocarcinoma risk: evidence based on 51106 subjects

OBJECTIVE

Several case-control studies have identified the association of the D919G polymorphism of the methionine synthase (MTR) gene with the risk of prostate adenocarcinoma (PRAD). However, the results were inconclusive.

MATERIALS AND METHODS

Odds ratios (ORs) with corresponding 95% confidence intervals (95% CIs) were evaluated to assess the correlation between MTR D919G variant and PRAD risk. In addition, in silico tools were used to demonstrate the relationship between MTR expression and PRAD risk and survival time.

RESULTS

The overall results from 10,617 PRAD cases and 40,489 control participants indicated the association of the MTR D919G variant with an increased risk of PRAD (allelic contrast: OR = 1.06, 95% CI = 1.01 - 1.11; GA vs. AA: OR = 1.08, 95% CI = 1.02 - 1.14; GG+GA vs. AA: OR = 1.08, 95% CI = 1.02 - 1.14). The stratified analysis yielded similar results for hospital based studies and those with larger sample sizes. Finally, the in silico results revealed lower MTR expression in PRAD tissue than in normal tissue (transcripts per million = 2.68 vs. 3.34, p<0.05). Furthermore, patients with high MTR expression and Gleason score = 6 exhibited reduced survival time (p<0.0001).

CONCLUSIONS

Our study indicated that the MTR D919G variant is associated with elevated risk to PRAD, especially for Asian descendants and hospital based studies. Moreover, the MTR D919G variant might be related to PRAD prognosis.

0802 To Examine the Effect of Gabapentin Enacarbil in Primary Restless Legs Syndrome Patients Who are on Dopaminergic Agents and Exhibiting Augmentation

Introduction

Augmentation is defined as worsening of the symptoms of Restless Legs Syndrome after a brief period of initial improvement with dopaminergic agents resulting in either an earlier onset, increase in severity, quicker onset, spread to other body parts. The exact prevalence of this phenomenon is not known and in patients experiencing augmentation, it can pose a difficult diagnostic and therapeutic challenge to the clinician. In our study, we found extended-release gabapentin to be an effective intervention in patients experiencing dopaminergic augmentation

Methods

This is an open-label single-arm study done in patients exhibiting augmentation while on dopaminergic agents. Patients who were enrolled in the study were initiated on oral extended-release gabapentin(Horizont) 600 mg at 5 pm at the beginning of the study. At day 90, attempts were made to reduce or discontinue dopaminergic agents. International Restless Legs Syndrome-Rating Scale (IRLS) and Augmentation Severity Rating Scale(ASRS) were recorded at each visit.

Results

A total of 10 patients were enrolled in the study while only 8 patients completed it. Compared to the baseline (visit 2), there is a significant improvement in both the augmentation severity(p= 0.0131) and the IRLS (p=0.0497). Wilcoxon matched-pairs signed rank test was used for statistical analysis.

Conclusion

Extended-release Gabapentin is an effective treatment option in primary RLS patients experiencing augmentation secondary to dopaminergic medication usage.

Support

The study is funded and medication is provided by Arbor Pharmaceuticals.

Implementing Mycoplasma genitalium testing across a London-based sexual health service: A quality improvement project

Recent national guidelines recommended testing for Mycoplasma genitalium (MG) in clinically-indicated conditions (CIC) including non-gonococcal urethritis (NGU), pelvic inflammatory disease (PID) and epididymo-orchitis. Over five months in 2018 a quality improvement project (QIP) was carried out across three London sexual health clinics with the aim of increasing MG testing rates in CICs. Three Plan-Do-Study-Act (PDSA) cycles were completed: improving IT access, an education event and reminder emails for clinicians who did not test in CIC. To measure testing rates ten patients from each CIC were randomly selected each week and MG testing outcomes were collected. As a balancing measure, we identified the rate of inappropriate MG testing. MG testing rates in patients with NGU increased to 90% following QIP initiation (baseline rate 60%) and this increase was sustained. No increase in MG testing was seen in PID and epididymo-orchitis. Inappropriate MG test rates were high (median of 11%) but remained constant throughout the QIP period. As MG testing is expanding across the UK, we outline a QIP integrating MG testing into a busy multi-site, sexual health service improving testing uptake while not increasing inappropriate testing.

Reconstruction and analysis of the aberrant lncRNA-miRNA-mRNA network in systemic lupus erythematosus

OBJECTIVE

A new perspective of determining the pathophysiology of systemic lupus erythematosus (SLE) development is required. The current study explores the aberrant expression of long non-coding RNAs (lncRNA), microRNA (miRNA) and mRNA. The study further constructs and analyses the lncRNA-miRNA-mRNA network to elucidate their gene regulation roles in SLE.

METHOD

We extracted mRNA, lncRNA and miRNA from the whole venous blood of 20 SLE patients and 20 normal control (NC) healthy individuals. A lncRNA-mRNA-miRNA network in SLE was constructed using a bioinformatics approach. Subsequently, Gene Ontology and Kyoto Encyclopedia of Genes and Genomes pathway analyses were performed using the Cytoscape plug-in BinGo, the DAVID database and Cytoscape software to explore the function of mRNAs in this network.

RESULT

A total of 855 mRNA, 7311 lncRNA and 134 miRNA with differentially expressed profiles were identified. Meanwhile, we established a competing endogenous RNA (ceRNA) subnetwork composed of 52 differentially expressed lncRNAs (DElncRNAs), seven differentially expressed miRNAs and 10 differentially expressed mRNAs. We extracted the subnetwork from the ceRNA network and found that three novel miRNAs were key: hsa-miR-145, hsa-miR-17 and hsa-miR-143. We also deduced that the DElncRNAs MIAT and NEAT1 might play crucial roles in the pathogenesis of SLE. The results were verified by bioinformatics analysis.

CONCLUSION

Our results provide a novel perspective for studying lncRNA-related and miRNA-related ceRNA networks in SLE.