Deep learning with robustness to missing data: A novel approach to the detection of COVID-19.

In the context of the current global pandemic and the limitations of the RT-PCR test, we propose a novel deep learning architecture, DFCN (Denoising Fully Connected Network). Since medical facilities around the world differ enormously in what laboratory tests or chest imaging may be available, DFCN is designed to be robust to missing input data. An ablation study extensively evaluates the performance benefits of the DFCN as well as its robustness to missing inputs. Data from 1088 patients with confirmed RT-PCR results are obtained from two independent medical facilities. The data includes results from 27 laboratory tests and a chest x-ray scored by a deep learning model. Training and test datasets are taken from different medical facilities. Data is made publicly available. The performance of DFCN in predicting the RT-PCR result is compared with 3 related architectures as well as a Random Forest baseline. All models are trained with varying levels of masked input data to encourage robustness to missing inputs. Missing data is simulated at test time by masking inputs randomly. DFCN outperforms all other models with statistical significance using random subsets of input data with 2-27 available inputs. When all 28 inputs are available DFCN obtains an AUC of 0.924, higher than any other model. Furthermore, with clinically meaningful subsets of parameters consisting of just 6 and 7 inputs respectively, DFCN achieves higher AUCs than any other model, with values of 0.909 and 0.919.

Rapid identification of SARS-CoV-2-infected patients at the emergency department using routine testing.

Objectives: The novel coronavirus disease 19 (COVID-19), caused by SARS-CoV-2, spreads rapidly across the world. The exponential increase in the number of cases has resulted in overcrowding of emergency departments (ED). Detection of SARS-CoV-2 is based on an RT-PCR of nasopharyngeal swab material. However, RT-PCR testing is time-consuming and many hospitals deal with a shortage of testing materials. Therefore, we aimed to develop an algorithm to rapidly evaluate an individual's risk of SARS-CoV-2 infection at the ED. Methods: In this multicenter retrospective study, routine laboratory parameters (C-reactive protein, lactate dehydrogenase, ferritin, absolute neutrophil and lymphocyte counts), demographic data and the chest X-ray/CT result from 967 patients entering the ED with respiratory symptoms were collected. Using these parameters, an easy-to-use point-based algorithm, called the corona-score, was developed to discriminate between patients that tested positive for SARS-CoV-2 by RT-PCR and those testing negative. Computational sampling was used to optimize the corona-score. Validation of the model was performed using data from 592 patients. Results: The corona-score model yielded an area under the receiver operating characteristic curve of 0.91 in the validation population. Patients testing negative for SARS-CoV-2 showed a median corona-score of 3 vs. 11 (scale 0-14) in patients testing positive for SARS-CoV-2 (p<0.001). Using cut-off values of 4 and 11 the model has a sensitivity and specificity of 96 and 95%, respectively. Conclusions: The corona-score effectively predicts SARS-CoV-2 RT-PCR outcome based on routine parameters. This algorithm provides the means for medical professionals to rapidly evaluate SARS-CoV-2 infection status of patients presenting at the ED with respiratory symptoms.

INR vs. thrombin generation assays for guiding VKA reversal: a retrospective comparison.

BACKGROUND: Prothrombin complex concentrate (PCC) is used to reverse vitamin K antagonist (VKA)-induced anticoagulation. Prothrombin time-derived international normalized ratio (INR) measurements are widely used in determining the required PCC dose, but this approach requires reappraisal. The aim of the present study was to determine the added value of the thrombin generation assay (TGA) compared with the INR in guidance of VKA reversal by PCC. METHODS: In an open, observational study, INR and TGA measurements were carried out on plasma samples from phenprocoumon-treated patients receiving VKA reversal. Following both analytical methods, PCC dosing correlates were calculated and compared retrospectively. Alternatively, in vitro PCC spiking experiments were performed. RESULTS: As expected, an exponential relationship between PCC dose and INR was found. For the TGA parameters peak thrombin and endogenous thrombin potential (ETP), however, this relationship was found to be linear throughout the full therapeutic range. Additional computational analysis showed a positive correlation (r²=0.7) between the initial INR and PCC dose required for a target INR of 2.1, which was completely lost at a lower target INR. In contrast, a positive correlation (r²=0.8) between initial ETP as well as peak height and PCC dose required to obtain parameter normalization was found. These correlates appeared useful for calculating PCC dose. CONCLUSIONS: Our results support the current debate questioning the rationale for the use of the INR in the management of anticoagulation by VKA. Compared with INR, TGA-based calculations may enable a more accurate PCC dosing regimen for patients requiring VKA reversal.

Endothelial cell-specific FGD5 involvement in vascular pruning defines neovessel fate in mice.

BACKGROUND: New vessel formation contributes to organ development during embryogenesis and tissue repair in response to mechanical damage, inflammation, and ischemia in adult organisms. Early angiogenesis includes formation of an excessive primitive network that needs to be reorganized into a secondary vascular network with higher hierarchical structure. Vascular pruning, the removal of aberrant neovessels by apoptosis, is a vital step in this process. Although multiple molecular pathways for early angiogenesis have been identified, little is known about the genetic regulators of secondary network development. METHODS AND RESULTS: Using a transcriptomics approach, we identified a new endothelial specific gene named FYVE, RhoGEF, and PH domain-containing 5 (FGD5) that plays a crucial role in vascular pruning. Loss- and gain-of-function studies demonstrate that FGD5 inhibits neovascularization, indicated by in vitro tube-formation, aortic-ring, and coated-bead assays and by in vivo coated-bead plug assays and studies in the murine retina model. FGD5 promotes apoptosis-induced vaso-obliteration via induction of the hey1-p53 pathway by direct binding and activation of cdc42. Indeed, FGD5 correlates with apoptosis in endothelial cells during vascular remodeling and was linked to rising p21(CIP1) levels in aging mice. CONCLUSION: We have identified FGD5 as a novel genetic regulator of vascular pruning by activation of endothelial cell-targeted apoptosis.

Role of delta-like-4/Notch in the formation and wiring of the lymphatic network in zebrafish.

OBJECTIVE: To study whether Notch signaling, which regulates cell fate decisions and vessel morphogenesis, controls lymphatic development. METHODS AND RESULTS: In zebrafish embryos, sprouts from the axial vein have lymphangiogenic potential because they give rise to the first lymphatics. Knockdown of delta-like-4 (Dll4) or its receptors Notch-1b or Notch-6 in zebrafish impaired lymphangiogenesis. Dll4/Notch silencing reduced the number of sprouts producing the string of parchordal lymphangioblasts; instead, sprouts connecting to the intersomitic vessels were formed. At a later phase, Notch silencing impaired navigation of lymphatic intersomitic vessels along their arterial templates. CONCLUSIONS: These studies imply critical roles for Notch signaling in the formation and wiring of the lymphatic network.

Vegfc/Flt4 signalling is suppressed by Dll4 in developing zebrafish intersegmental arteries.

The development of arteries, veins and lymphatics from pre-existing vessels are intimately linked processes controlled by a number of well-studied reiteratively acting signalling pathways. To delineate the mechanisms governing vessel formation in vivo, we performed a forward genetic screen in zebrafish and isolated the mutant expando. Molecular characterisation revealed a loss-of-function mutation in the highly conserved kinase insert region of flt4. Consistent with previous reports, flt4 mutants were deficient in lymphatic vascular development. Recent studies have demonstrated a role for Flt4 in blood vessels and showed that Dll4 limits angiogenic potential by limiting Flt4 function in developing blood vessels. We found that arterial angiogenesis proceeded normally, yet the dll4 loss-of-function arterial hyperbranching phenotype was rescued, in flt4 signalling mutants. Furthermore, we found that the Flt4 ligand Vegfc drives arterial hyperbranching in the absence of dll4. Upon knockdown of dll4, intersegmental arteries were sensitised to increased vegfc levels and the overexpression of dll4 inhibited Vegfc/Flt4-dependent angiogenesis events. Taken together, these data demonstrate that dll4 functions to suppress the ability of developing intersegmental arteries to respond to Vegfc-driven Flt4 signalling in zebrafish. We propose that this mechanism contributes to the differential response of developing arteries and veins to a constant source of Vegfc present in the embryo during angiogenesis.

Redundant roles for sox7 and sox18 in arteriovenous specification in zebrafish.

The specification of arteries and veins is an essential process in establishing and maintaining a functional blood vessel system. Incorrect arteriovenous specification disrupts embryonic development but has also been diagnosed in human syndromes such as hypotrichosis-lymphedema-telangiectasia, characterized by defects in blood and lymphatic vessels and associated with mutations in SOX18. Here we characterize the role of sox7 and sox18 during zebrafish vasculogenesis. Sox7 and sox18 are specifically expressed in the developing vasculature, and simultaneous loss of their function results in a severe loss of the arterial identity of the presumptive aorta which instead expresses venous markers, followed by dramatic arteriovenous shunt formations. Our study identifies members of the Sox family as key factors in specifying arteriovenous identity and will help to better understand hypotrichosis-lymphedema-telangiectasia and other diseases.

Stereoselectivity and substrate specificity in the kinetic resolution of methyl-substituted 1-oxaspiro[2.5]octanes by Rhodotorula glutinis epoxide hydrolase.

The kinetic resolution of a range of methyl-substituted 1-oxaspiro[2.5]octanes by yeast epoxide hydrolase (YEH) from Rhodotorula glutinis has been investigated. The structural determinants of substrate specificity and stereoselectivity of YEH toward these substrates appeared to be the configuration of the epoxide ring and the substitution pattern of the cyclohexane ring. For all compounds tested, O-axial epoxides were hydrolyzed faster than the corresponding O-equatorial compounds. In concern of the ring substituents, YEH preferred methyl groups on the Re side of the ring. Placement of substituents close to the spiroepoxide carbon decreased the reaction rate but increased enantioselectivity. YEH-catalyzed kinetic resolutions of 4-methyl 1-oxaspiro[2.5]octane epimers were most enantioselective (E > 100).