Structure-Guided Design of an Anti-dengue Antibody Directed to a Non-immunodominant Epitope.

Dengue is the most common vector-borne viral disease, causing nearly 400 million infections yearly. Currently there are no approved therapies. Antibody epitopes that elicit weak humoral responses may not be accessible by conventional B cell panning methods. To demonstrate an alternative strategy to generating a therapeutic antibody, we employed a non-immunodominant, but functionally relevant, epitope in domain III of the E protein, and engineered by structure-guided methods an antibody directed to it. The resulting antibody, Ab513, exhibits high-affinity binding to, and broadly neutralizes, multiple genotypes within all four serotypes. To assess therapeutic relevance of Ab513, activity against important human clinical features of dengue was investigated. Ab513 mitigates thrombocytopenia in a humanized mouse model, resolves vascular leakage, reduces viremia to nearly undetectable levels, and protects mice in a maternal transfer model of lethal antibody-mediated enhancement. The results demonstrate that Ab513 may reduce the public health burden from dengue.

Structural determinants for naturally evolving H5N1 hemagglutinin to switch its receptor specificity.

Of the factors governing human-to-human transmission of the highly pathogenic avian-adapted H5N1 virus, the most critical is the acquisition of mutations on the viral hemagglutinin (HA) to "quantitatively switch" its binding from avian to human glycan receptors. Here, we describe a structural framework that outlines a necessary set of H5 HA receptor-binding site (RBS) features required for the H5 HA to quantitatively switch its preference to human receptors. We show here that the same RBS HA mutations that lead to aerosol transmission of A/Vietnam/1203/04 and A/Indonesia/5/05 viruses, when introduced in currently circulating H5N1, do not lead to a quantitative switch in receptor preference. We demonstrate that HAs from circulating clades require as few as a single base pair mutation to quantitatively switch their binding to human receptors. The mutations identified by this study can be used to monitor the emergence of strains having human-to-human transmission potential.

Glycan receptor binding of the influenza A virus H7N9 hemagglutinin.

The advent of H7N9 in early 2013 is of concern for a number of reasons, including its capability to infect humans, the lack of clarity in the etiology of infection, and because the human population does not have pre-existing immunity to the H7 subtype. Earlier sequence analyses of H7N9 hemagglutinin (HA) point to amino acid changes that predicted human receptor binding and impinge on the antigenic characteristics of the HA. Here, we report that the H7N9 HA shows limited binding to human receptors; however, should a single amino acid mutation occur, this would result in structural changes within the receptor binding site that allow for extensive binding to human receptors present in the upper respiratory tract. Furthermore, a subset of the H7N9 HA sequences demarcating coevolving amino acids appears to be in the antigenic regions of H7, which, in turn, could impact effectiveness of the current WHO-recommended prepandemic H7 vaccines.

Development of an antibody fused with an antimicrobial peptide targeting Pseudomonas aeruginosa: A new approach to prevent and treat bacterial infections.

The increase in emerging drug resistant Gram-negative bacterial infections is a global concern. In addition, there is growing recognition that compromising the microbiota through the use of broad-spectrum antibiotics can impact long term patient outcomes. Therefore, there is the need to develop new bactericidal strategies to combat Gram-negative infections that would address these specific issues. In this study, we report and characterize one such approach, an antibody-drug conjugate (ADC) that combines (i) targeting the surface of a specific pathogenic organism through a monoclonal antibody with (ii) the high killing activity of an antimicrobial peptide. We focused on a major pathogenic Gram-negative bacterium associated with antibacterial resistance: Pseudomonas aeruginosa. To target this organism, we designed an ADC by fusing an antimicrobial peptide to the C-terminal end of the VH and/or VL-chain of a monoclonal antibody, VSX, that targets the core of P. aeruginosa lipopolysaccharide. This ADC demonstrates appropriately minimal levels of toxicity against mammalian cells, rapidly kills P. aeruginosa strains, and protects mice from P. aeruginosa lung infection when administered therapeutically. Furthermore, we found that the ADC was synergistic with several classes of antibiotics. This approach described in this study might result in a broadly useful strategy for targeting specific pathogenic microorganisms without further augmenting antibiotic resistance.

Characterization of an Antibody Recognizing the Conserved Inner Core of Pseudomonas aeruginosa Lipopolysaccharides.

Bacterial infections are a growing public health threat with carbapenem-resistant Pseudomonas aeruginosa being classified as a Priority 1 critical threat by the World Health Organization. Antibody-based therapeutics can serve as an alternative and in some cases supplement antibiotics for the treatment of bacterial infections. The glycans covering the bacterial cell surface have been proposed as intriguing targets for binding by antibodies; however, antibodies that can engage with high affinity and specificity with glycans are much less common compared to antibodies that engage with protein antigens. In this study, we sought to characterize an antibody that targets a conserved glycan epitope on the surface of Pseudomonas. First, we characterized the breadth of binding of VSX, demonstrating that the VSX is specific to Pseudomonas but can bind across multiple serotypes of the organism. Next, we provide insight into how VSX engages with its target epitope, using a combination of biolayer interferometry and nuclear magnetic resonance, and verify our results using site-directed mutagenesis experiments. We demonstrate that the antibody, with limited somatic hypermutation of the complementarity-determining regions (CDRs) and with a characteristic set of arginines within the CDRs, specifically targets the conserved inner core of Pseudomonas lipopolysaccharides. Our results provide important additional context to antibody-glycan contacts and provide insight useful for the construction of vaccines and therapeutics against Pseudomonas aeruginosa, an important human pathogen.

Structural prediction of antibody-APRIL complexes by computational docking constrained by antigen saturation mutagenesis library data.

IgA nephropathy (IgAN) is the most prevalent cause of primary glomerular disease worldwide, and the cytokine A PRoliferation-Inducing Ligand (APRIL) is emerging as a key player in IgAN pathogenesis and disease progression. For a panel of anti-human APRIL antibodies with known antagonistic activity, we sought to define their structural mode of engagement to understand molecular mechanisms of action and aid rational antibody engineering. Reliable computational prediction of antibody-antigen complexes remains challenging, and experimental methods such as X-ray co-crystallography and cryoEM have considerable technical, resource, and throughput barriers. To overcome these limitations, we implemented an integrated and accessible experimental-computational workflow to more accurately predict structures of antibody-APRIL complexes. Specifically, a yeast surface display library encoding site-saturation mutagenized surface positions of APRIL was screened against a panel of anti-APRIL antibodies to rapidly obtain a comprehensive biochemical profile of mutational impact on binding for each antibody. The experimentally derived mutational profile data were used as quantitative constraints in a computational docking workflow optimized for antibodies, resulting in robust structural models of antibody-antigen complexes. The model results were confirmed by solving the cocrystal structure of one antibody-APRIL complex, which revealed strong agreement with our model. The models were used to rationally select and engineer one antibody for cross-species APRIL binding, which quite often aids further testing in relevant animal models. Collectively, we demonstrate a rapid, integrated computational-experimental approach to robustly predict antibody-antigen structures information, which can aid rational antibody engineering and provide insights into molecular mechanisms of action.

Antibody-Bactericidal Macrocyclic Peptide Conjugates To Target Gram-Negative Bacteria.

To combat antimicrobial infections, new active molecules are needed. Antimicrobial peptides, ever abundant in nature, are a fertile starting point to develop new antimicrobial agents but suffer from low stability, low specificity, and off-target toxicity. These drawbacks have limited their development. To overcome some of these limitations, we developed antibody-bactericidal macrocyclic peptide conjugates (ABCs), in which the antibody directs the bioactive macrocyclic peptide to the targeted Gram-negative bacteria. We used cysteine SN Ar chemistry to synthesize and systematically study a library of large (>30-mer) macrocyclic antimicrobial peptides (mAMPs) to discover variants with extended proteolytic stability in human serum and low hemolytic activity while maintaining bioactivity. We then conjugated, by using sortase A, these bioactive variants onto an Escherichia coli targeted monoclonal antibody. We found that these ABCs had minimized hemolytic activity and were able to kill E. coli at nanomolar concentrations. Our findings suggest macrocyclic peptides if fused to antibodies may facilitate the discovery of new agents to treat bacterial infections.

A broadly neutralizing human monoclonal antibody is effective against H7N9.

Emerging strains of influenza represent a significant public health threat with potential pandemic consequences. Of particular concern are the recently emerged H7N9 strains which cause pneumonia with acute respiratory distress syndrome. Estimates are that nearly 80% of hospitalized patients with H7N9 have received intensive care unit support. VIS410, a human antibody, targets a unique conserved epitope on influenza A. We evaluated the efficacy of VIS410 for neutralization of group 2 influenza strains, including H3N2 and H7N9 strains in vitro and in vivo. VIS410, administered at 50 mg/kg, protected DBA mice infected with A/Anhui/2013 (H7N9), resulting in significant survival benefit upon single-dose (-24 h) or double-dose (-12 h, +48 h) administration (P < 0.001). A single dose of VIS410 at 50 mg/kg (-12 h) combined with oseltamivir at 50 mg/kg (-12 h, twice daily for 7 d) in C57BL/6 mice infected with A/Shanghai 2/2013 (H7N9) resulted in significant decreased lung viral load (P = 0.002) and decreased lung cytokine responses for nine of the 11 cytokines measured. Based on these results, we find that VIS410 may be effective either as monotherapy or combined with antivirals in treating H7N9 disease, as well as disease from other influenza strains.

Rapid Device-Detected Nonsustained Ventricular Tachycardia in the Risk Stratification of Hypertrophic Cardiomyopathy.

BACKGROUND: Nonsustained ventricular tachycardia (NSVT) detected by ambulatory Holter (Holter NSVT) is a major risk factor for sudden cardiac death in hypertrophic cardiomyopathy (HCM). We hypothesized that the prognostic utility of Holter NSVT in HCM would improve with prolonged monitoring and a higher heart rate cut-off for detection. METHODS: We enrolled 60 patients (44 ± 14 years) with HCM, who had a prophylactic implantable cardioverter defibrillator (ICD). Positive Holter NSVT (prior to implant) was defined as ≥3 beats at ≥120 beats per minute (bpm). We assessed the prevalence of rapid NSVT (RNSVT) detected by their ICD within 12 months of its implant, defined as 4-16 beats at ≥150-200 bpm. The primary outcome was appropriate ICD therapy (antitachycardia pacing and shocks) for sustained ventricular arrhythmia (VA). RESULTS: Holter NSVT was detected in 34 patients. RNSVT occurred in 21 (35%) patients of whom five did not have Holter NSVT. Over a median follow-up of 61 (interquartile range 29, 129) months after ICD implant, nine patients had VA. RNSVT, but not Holter NSVT, was significantly associated with VA (hazard ratio 6.2, 95% confidence interval [1.3-30], P = 0.01) by multivariable Cox regression analysis that included conventional risk factors. Receiver operating characteristic analysis for RNSVT (area under curve 0.80, P = 0.005) showed that the occurrence of ≥2 episodes of RNSVT discriminated patients for VA optimally (sensitivity 78%, specificity 84%, positive predictive value 47%, negative predictive value 96%). CONCLUSIONS: In this pilot study, RNSVT detected by continuous monitoring independently predicted VA in HCM and offered superior discrimination of VA risk compared to conventional risk factors, including Holter NSVT. Future studies are needed to validate these findings in a larger, unselected HCM cohort.

Insights into the human glycan receptor conformation of 1918 pandemic hemagglutinin-glycan complexes derived from nuclear magnetic resonance and molecular dynamics studies.

The glycan receptor binding and specificity of influenza A viral hemagglutinin (HA) are critical for virus infection and transmission in humans. However, ambiguities in the interpretation of the receptor binding specificity of hemagglutinin from human- and avian-adapted viruses have prevented an understanding of its relationship with aerosol transmissibility, an exclusive property of human-adapted viruses. A previous conformational study, which we performed, indicated that human and avian receptors sample distinct conformations in solution. On the basis of detailed nuclear magnetic resonance (NMR) studies provided herein, we offer evidence of the distinct structural constraints imposed by hemagglutinin receptor binding sites on the glycan conformational space upon binding. The hemagglutinin from the SC18 virus, which has efficient aerosol transmissibility in humans (human-adapted), imposed the most stringent constraints on the conformational space of the human glycan receptor (LSTc), compared to single (NY18) or double (AV18) amino acid HA mutants, a property correlating to the ligand-HA binding strength. This relationship was also observed for the avian-adapted HA, where the high affinity binding partner, AV18, imposed the most stringent conformational constraints on the avian receptor, compared to those imposed by NY18. In particular, it is interesting to observe how different HAs when binding to human or avian glycosidic receptors impose significantly different conformational states, in terms of the states sampled by the glycosidic backbone and/or the entire molecule shape (linear or bent), when compared to the corresponding unbound glycans. Significantly, we delineate a "characteristic NMR signature" for the human adapted hemagglutinin (SC18) binding to human glycan receptors. Therefore, the conformational space constraints imposed by the hemagglutinin receptor binding site provide a characteristic signature that could be a useful tool for the surveillance of human adaptation of other (such as H7N9 and H5N1) deadly influenza viruses.

Sialylneolacto-N-tetraose c (LSTc)-bearing liposomal decoys capture influenza A virus.

Influenza is a severe disease in humans and animals with few effective therapies available. All strains of influenza virus are prone to developing drug resistance due to the high mutation rate in the viral genome. A therapeutic agent that targets a highly conserved region of the virus could bypass resistance and also be effective against multiple strains of influenza. Influenza uses many individually weak ligand binding interactions for a high avidity multivalent attachment to sialic acid-bearing cells. Polymerized sialic acid analogs can form multivalent interactions with influenza but are not ideal therapeutics due to solubility and toxicity issues. We used liposomes as a novel means for delivery of the glycan sialylneolacto-N-tetraose c (LSTc). LSTc-bearing decoy liposomes form multivalent, polymer-like interactions with influenza virus. Decoy liposomes competitively bind influenza virus in hemagglutination inhibition assays and inhibit infection of target cells in a dose-dependent manner. Inhibition is specific for influenza virus, as inhibition of Sendai virus and respiratory syncytial virus is not observed. In contrast, monovalent LSTc does not bind influenza virus or inhibit infectivity. LSTc decoy liposomes prevent the spread of influenza virus during multiple rounds of replication in vitro and extend survival of mice challenged with a lethal dose of virus. LSTc decoy liposomes co-localize with fluorescently tagged influenza virus, whereas control liposomes do not. Considering the conservation of the hemagglutinin binding pocket and the ability of decoy liposomes to form high avidity interactions with influenza hemagglutinin, our decoy liposomes have potential as a new therapeutic agent against emerging influenza strains.

Immunogenicity and protective efficacy of a major White Spot Syndrome Virus (WSSV) envelope protein VP24 expressed in Escherichia coli against WSSV.

The study reports cloning, expression and characterization of immunogenic activity of VP24, a major envelope protein of White Spot Syndrome Virus (WSSV). His-tagged VP24 was expressed as truncated protein and purified from inclusion bodies by metal affinity chromatography under denaturing conditions. The ability to confer protection from WSSV by oral administration of recombinant viral protein (rVP24) was examined in black tiger shrimp Penaeus monodon (P. monodon) juveniles (advanced post larvae). Animals were fed with rVP24 for 10 days, orally challenged with WSSV and assayed for expression of viral genes and shrimp immune genes on the 2nd, 5th and 8th days of challenge. The survival of juvenile shrimps in the vaccinated and challenged group was significantly higher compared to the unvaccinated and challenged group with lesser viral gene expression (DNA polymerase, latency 1 and vp28). Analysis of immune gene expression showed upregulation of syntenin and down regulation of STAT, Rab 7 and caspase during the experimental period. This study points to the feasibility of using rVP24 as candidate vaccine in P. monodon against WSSV.

Training and Comparison of nnU-Net and DeepMedic Methods for Autosegmentation of Pediatric Brain Tumors.

BACKGROUND AND PURPOSE: Tumor segmentation is essential in surgical and treatment planning and response assessment and monitoring in pediatric brain tumors, the leading cause of cancer-related death among children. However, manual segmentation is time-consuming and has high interoperator variability, underscoring the need for more efficient methods. After training, we compared 2 deep-learning-based 3D segmentation models, DeepMedic and nnU-Net, with pediatric-specific multi-institutional brain tumor data based on multiparametric MR images. MATERIALS AND METHODS: Multiparametric preoperative MR imaging scans of 339 pediatric patients (n = 293 internal and n = 46 external cohorts) with a variety of tumor subtypes were preprocessed and manually segmented into 4 tumor subregions, ie, enhancing tumor, nonenhancing tumor, cystic components, and peritumoral edema. After training, performances of the 2 models on internal and external test sets were evaluated with reference to ground truth manual segmentations. Additionally, concordance was assessed by comparing the volume of the subregions as a percentage of the whole tumor between model predictions and ground truth segmentations using the Pearson or Spearman correlation coefficients and the Bland-Altman method. RESULTS: The mean Dice score for nnU-Net internal test set was 0.9 (SD, 0.07) (median, 0.94) for whole tumor; 0.77 (SD, 0.29) for enhancing tumor; 0.66 (SD, 0.32) for nonenhancing tumor; 0.71 (SD, 0.33) for cystic components, and 0.71 (SD, 0.40) for peritumoral edema, respectively. For DeepMedic, the mean Dice scores were 0.82 (SD, 0.16) for whole tumor; 0.66 (SD, 0.32) for enhancing tumor; 0.48 (SD, 0.27) for nonenhancing tumor; 0.48 (SD, 0.36) for cystic components, and 0.19 (SD, 0.33) for peritumoral edema, respectively. Dice scores were significantly higher for nnU-Net (P ≤ .01). Correlation coefficients for tumor subregion percentage volumes were higher (0.98 versus 0.91 for enhancing tumor, 0.97 versus 0.75 for nonenhancing tumor, 0.98 versus 0.80 for cystic components, 0.95 versus 0.33 for peritumoral edema in the internal test set). Bland-Altman plots were better for nnU-Net compared with DeepMedic. External validation of the trained nnU-Net model on the multi-institutional Brain Tumor Segmentation Challenge in Pediatrics (BraTS-PEDs) 2023 data set revealed high generalization capability in the segmentation of whole tumor, tumor core (a combination of enhancing tumor, nonenhancing tumor, and cystic components), and enhancing tumor with mean Dice scores of 0.87 (SD, 0.13) (median, 0.91), 0.83 (SD, 0.18) (median, 0.89), and 0.48 (SD, 0.38) (median, 0.58), respectively. CONCLUSIONS: The pediatric-specific data-trained nnU-Net model is superior to DeepMedic for whole tumor and subregion segmentation of pediatric brain tumors.

Modular glycosphere assays for high-throughput functional characterization of influenza viruses.

BACKGROUND: The ongoing global efforts to control influenza epidemics and pandemics require high-throughput technologies to detect, quantify, and functionally characterize viral isolates. The 2009 influenza pandemic as well as the recent in-vitro selection of highly transmissible H5N1 variants have only increased existing concerns about emerging influenza strains with significantly enhanced human-to-human transmissibility. High-affinity binding of the virus hemagglutinin to human receptor glycans is a highly sensitive and stringent indicator of host adaptation and virus transmissibility. The surveillance of receptor-binding characteristics can therefore provide a strong additional indicator for the relative hazard imposed by circulating and newly emerging influenza strains. RESULTS: Streptavidin-coated microspheres were coated with selected biotinylated glycans to mimic either human or avian influenza host-cell receptors. Such glycospheres were used to selectively capture influenza virus of diverse subtypes from a variety of samples. Bound virus was then detected by fluorescently labelled antibodies and analyzed by quantitative flow cytometry. Recombinant hemagglutinin, inactivated virus, and influenza virions were captured and analyzed with regards to receptor specificity over a wide range of analyte concentration. High-throughput analyses of influenza virus produced dose-response curves that allow for functional assessment of relative receptor affinity and thus transmissibility. CONCLUSIONS: Modular glycosphere assays for high-throughput functional characterization of influenza viruses introduce an important tool to augment the surveillance of clinical and veterinarian influenza isolates with regards to receptor specificity, host adaptation, and virus transmissibility.

Glycosylation at Asn91 of H1N1 haemagglutinin affects binding to glycan receptors.

The glycoprotein HA (haemagglutinin) on the surface of influenza A virus plays a central role in recognition and binding to specific host cell-surface glycan receptors and in fusion of viral membrane to the host nuclear membrane during viral replication. Given the abundance of HA on the viral surface, this protein is also the primary target for host innate and adaptive immune responses. Although addition of glycosylation sites on HA are a part of viral evolution to evade the host immune responses, there are specific glycosylation sites that are conserved during most of the evolution of the virus. In the present study, it was demonstrated that one such conserved glycosylation site at Asn(91) in H1N1 HA critically governs the glycan receptor-binding specificity and hence would potentially impinge on the host adaptation of the virus.

Streamlining pipeline efficiency: a novel model-agnostic technique for accelerating conditional generative and virtual screening pipelines.

The discovery of potential therapeutic agents for life-threatening diseases has become a significant problem. There is a requirement for fast and accurate methods to identify drug-like molecules that can be used as potential candidates for novel targets. Existing techniques like high-throughput screening and virtual screening are time-consuming and inefficient. Traditional molecule generation pipelines are more efficient than virtual screening but use time-consuming docking software. Such docking functions can be emulated using Machine Learning models with comparable accuracy and faster execution times. However, we find that when pre-trained machine learning models are employed in generative pipelines as oracles, they suffer from model degradation in areas where data is scarce. In this study, we propose an active learning-based model that can be added as a supplement to enhanced molecule generation architectures. The proposed method uses uncertainty sampling on the molecules created by the generator model and dynamically learns as the generator samples molecules from different regions of the chemical space. The proposed framework can generate molecules with high binding affinity with [Formula: see text]a 70% improvement in runtime compared to the baseline model by labeling only [Formula: see text]30% of molecules compared to the baseline oracle.

Automated Tumor Segmentation and Brain Tissue Extraction from Multiparametric MRI of Pediatric Brain Tumors: A Multi-Institutional Study.

Background: Brain tumors are the most common solid tumors and the leading cause of cancer-related death among all childhood cancers. Tumor segmentation is essential in surgical and treatment planning, and response assessment and monitoring. However, manual segmentation is time-consuming and has high interoperator variability. We present a multi-institutional deep learning-based method for automated brain extraction and segmentation of pediatric brain tumors based on multi-parametric MRI scans. Methods: Multi-parametric scans (T1w, T1w-CE, T2, and T2-FLAIR) of 244 pediatric patients (n=215 internal and n=29 external cohorts) with de novo brain tumors, including a variety of tumor subtypes, were preprocessed and manually segmented to identify the brain tissue and tumor subregions into four tumor subregions, i.e., enhancing tumor (ET), non-enhancing tumor (NET), cystic components (CC), and peritumoral edema (ED). The internal cohort was split into training (n=151), validation (n=43), and withheld internal test (n=21) subsets. DeepMedic, a three-dimensional convolutional neural network, was trained and the model parameters were tuned. Finally, the network was evaluated on the withheld internal and external test cohorts. Results: Dice similarity score (median±SD) was 0.91±0.10/0.88±0.16 for the whole tumor, 0.73±0.27/0.84±0.29 for ET, 0.79±19/0.74±0.27 for union of all non-enhancing components (i.e., NET, CC, ED), and 0.98±0.02 for brain tissue in both internal/external test sets. Conclusions: Our proposed automated brain extraction and tumor subregion segmentation models demonstrated accurate performance on segmentation of the brain tissue and whole tumor regions in pediatric brain tumors and can facilitate detection of abnormal regions for further clinical measurements. Key Points: We proposed automated tumor segmentation and brain extraction on pediatric MRI.The volumetric measurements using our models agree with ground truth segmentations. Importance of the Study: The current response assessment in pediatric brain tumors (PBTs) is currently based on bidirectional or 2D measurements, which underestimate the size of non-spherical and complex PBTs in children compared to volumetric or 3D methods. There is a need for development of automated methods to reduce manual burden and intra- and inter-rater variability to segment tumor subregions and assess volumetric changes. Most currently available automated segmentation tools are developed on adult brain tumors, and therefore, do not generalize well to PBTs that have different radiological appearances. To address this, we propose a deep learning (DL) auto-segmentation method that shows promising results in PBTs, collected from a publicly available large-scale imaging dataset (Children's Brain Tumor Network; CBTN) that comprises multi-parametric MRI scans of multiple PBT types acquired across multiple institutions on different scanners and protocols. As a complementary to tumor segmentation, we propose an automated DL model for brain tissue extraction.

Automated tumor segmentation and brain tissue extraction from multiparametric MRI of pediatric brain tumors: A multi-institutional study.

Background: Brain tumors are the most common solid tumors and the leading cause of cancer-related death among all childhood cancers. Tumor segmentation is essential in surgical and treatment planning, and response assessment and monitoring. However, manual segmentation is time-consuming and has high interoperator variability. We present a multi-institutional deep learning-based method for automated brain extraction and segmentation of pediatric brain tumors based on multi-parametric MRI scans. Methods: Multi-parametric scans (T1w, T1w-CE, T2, and T2-FLAIR) of 244 pediatric patients ( n = 215 internal and n = 29 external cohorts) with de novo brain tumors, including a variety of tumor subtypes, were preprocessed and manually segmented to identify the brain tissue and tumor subregions into four tumor subregions, i.e., enhancing tumor (ET), non-enhancing tumor (NET), cystic components (CC), and peritumoral edema (ED). The internal cohort was split into training ( n = 151), validation ( n = 43), and withheld internal test ( n = 21) subsets. DeepMedic, a three-dimensional convolutional neural network, was trained and the model parameters were tuned. Finally, the network was evaluated on the withheld internal and external test cohorts. Results: Dice similarity score (median ± SD) was 0.91 ± 0.10/0.88 ± 0.16 for the whole tumor, 0.73 ± 0.27/0.84 ± 0.29 for ET, 0.79 ± 19/0.74 ± 0.27 for union of all non-enhancing components (i.e., NET, CC, ED), and 0.98 ± 0.02 for brain tissue in both internal/external test sets. Conclusions: Our proposed automated brain extraction and tumor subregion segmentation models demonstrated accurate performance on segmentation of the brain tissue and whole tumor regions in pediatric brain tumors and can facilitate detection of abnormal regions for further clinical measurements.

A multi-institutional pediatric dataset of clinical radiology MRIs by the Children's Brain Tumor Network.

Pediatric brain and spinal cancers remain the leading cause of cancer-related death in children. Advancements in clinical decision-support in pediatric neuro-oncology utilizing the wealth of radiology imaging data collected through standard care, however, has significantly lagged other domains. Such data is ripe for use with predictive analytics such as artificial intelligence (AI) methods, which require large datasets. To address this unmet need, we provide a multi-institutional, large-scale pediatric dataset of 23,101 multi-parametric MRI exams acquired through routine care for 1,526 brain tumor patients, as part of the Children's Brain Tumor Network. This includes longitudinal MRIs across various cancer diagnoses, with associated patient-level clinical information, digital pathology slides, as well as tissue genotype and omics data. To facilitate downstream analysis, treatment-naïve images for 370 subjects were processed and released through the NCI Childhood Cancer Data Initiative via the Cancer Data Service. Through ongoing efforts to continuously build these imaging repositories, our aim is to accelerate discovery and translational AI models with real-world data, to ultimately empower precision medicine for children.

Automated Quantification of Abnormal QRS Peaks From High-Resolution ECGs Predicts Late Ventricular Arrhythmias in Hypertrophic Cardiomyopathy: A 5-Year Prospective Multicenter Study.

Background Patients with hypertrophic cardiomyopathy (HCM) are at risk of ventricular arrhythmia (VA) attributed to abnormal electrical activation arising from myocardial fibrosis and myocyte disarray. We sought to quantify intra-QRS peaks (QRSp) in high-resolution ECGs as a measure of abnormal activation to predict late VA in patients with HCM. Methods and Results Prospectively enrolled patients with HCM (n=143, age 53±14 years) with prophylactic implantable cardioverter-defibrillators had 3-minute, high-resolution (1024 Hz), digital 12-lead ECGs recorded during intrinsic rhythm. For each precordial lead, QRSp was defined as the total number of peaks detected in the QRS complex that deviated from a smoothing filtered version of the QRS. The VA end point was appropriate implantable cardioverter-defibrillator therapy during 5-year prospective follow-up. After 5 years, 21 (16%) patients had VA. Patients who were VA positive had greater QRSp (6.0 [4.0-7.0] versus 4.0 [2.0-5.0]; P<0.01) and lower left ventricular ejection fraction (57±11 versus 62±9; P=0.038) compared with patients who were VA negative, but had similar established HCM risk metrics. Receiver operating characteristic analysis revealed that QRSp discriminated VA (area under the curve=0.76; P<0.001), with a QRSp ≥4 achieving 91% sensitivity and 39% specificity. The annual VA rate was greater in patients with QRSp ≥4 versus QRSp <4 (4.4% versus 0.98%; P=0.012). In multivariable Cox regression, age <50 years (hazard ratio [HR], 2.53; P=0.009) and QRSp (HR per QRS peak, 1.41; P=0.009) predicted VA after adjusting for established HCM risk metrics. In patients aged <50 years, the annual VA rate was 0.0% for QRSp <4 compared with 6.9% for QRSp ≥4 (P=0.012). Conclusions QRSp predicted VA in patients with HCM who were eligible for an implantable cardioverter-defibrillator after adjusting for established HCM risk metrics, such that each additional QRS peak increases VA risk by 40%. QRSp <4 was associated with a <1% annual VA risk in all patients, and no VA risk among those aged <50 years. This novel ECG metric may improve patient selection for prophylactic implantable cardioverter-defibrillator therapy by identifying those with low VA risk. These findings require further validation in a lower risk HCM cohort. Registration URL: https://www.clinicaltrials.gov; Unique identifier: NCT02560844.