Search

Meningococcal ( Neisseria meningitidis ) serogroup B (MenB) strain antigens are diverse and a limited number of strains can be evaluated using the human serum bactericidal antibody (hSBA) assay. The genetic Meningococcal Antigen Typing System (gMATS) was developed to predict the likelihood of coverage for large numbers of isolates by the 4CMenB vaccine, which includes antigens Neisseria adhesin A (NadA), Neisserial Heparin-Binding Antigen (NHBA), factor H-binding protein (fHbp), and Porin A (PorA). In this study, we characterized by whole-genome analyses 284 invasive MenB isolates collected from 2010 to 2014 by the Argentinian National Laboratories Network (52-61 isolates per year). Strain coverage was estimated by gMATS on all isolates and by hSBA assay on 74 randomly selected isolates, representative of the whole panel. The four most common clonal complexes (CCs), accounting for 81.3% of isolates, were CC-865 (75 isolates, 26.4%), CC-32 (59, 20.8%), CC-35 (59, 20.8%), and CC-41/44 (38, 13.4%). Vaccine antigen genotyping showed diversity. The most prevalent variants/peptides were fHbp variant 2, NHBA peptides 24, 21, and 2, and PorA variable region 2 profiles 16-36 and 14. The nadA gene was present in 66 (23.2%) isolates. Estimated strain coverage by hSBA assay showed 78.4% of isolates were killed by pooled adolescent sera, and 51.4% and 64.9% (based on two different thresholds) were killed by pooled infant sera. Estimated coverage by gMATS (61.3%; prediction interval: 55.5%, 66.7%) was consistent with the infant hSBA assay results. Continued genomic surveillance is needed to evaluate the persistence of major MenB CCs in Argentina.

Whole genome sequencing of bacteria is important to enable strain classification. Using entire genomes as an input to machine learning (ML) models would allow rapid classification of strains while using information from multiple genetic elements. We developed a "bag-of-words" approach to encode, using SentencePiece or k-mer tokenization, entire bacterial genomes and analyze these with ML. Initial model selection identified SentencePiece with 8,000 and 32,000 words as the best approach for genome tokenization. We then classified in Neisseria meningitidis genomes the capsule B group genotype with 99.6% accuracy and the multifactor invasive phenotype with 90.2% accuracy, in an independent test set. Subsequently, in silico knockouts of 2,808 genes confirmed that the ML model predictions aligned with our current understanding of the underlying biology. To our knowledge, this is the first ML method using entire bacterial genomes to classify strains and identify genes considered relevant by the classifier., Competing Interests: M.P. and S.B. disclose that their postdoctoral grant at the University of Pisa, Italy, is funded by GSK. Outside of the submitted work, S.B. also discloses having received grants from the University of Siena, Italy and the University of Tuscia, Viterbo, Italy; payment or honoraria for lectures, presentations, speakers’ bureaus, manuscript writing or educational events from the University of Siena, Italy, University of Bari, Italy, and the Italica Academy srl. A.P., A.B., G.R., A.M., and M.B. are employed by GSK. M.S. was an intern at GSK during the time of the study. G.R. holds shares in GSK and in Novartis AG. A.M. holds shares in GSK. C.P. and A.S. disclose that GSK commissioned this research. The authors declare no other financial and non-financial relationships and activities and no other conflicts of interest., (© 2024 GSK.)

Risk and outcome of acute promyelocytic leukemia (APL) are particularly worsened in obese-overweight individuals, but the underlying molecular mechanism is unknown. In established mouse APL models (Ctsg-PML::RARA), we confirmed that obesity induced by high-fat diet (HFD) enhances leukemogenesis by increasing penetrance and shortening latency, providing an ideal model to investigate obesity-induced molecular events in the preleukemic phase. Surprisingly, despite increasing DNA damage in hematopoietic stem cells (HSC), HFD only minimally increased mutational load, with no relevant impact on known cancer-driving genes. HFD expanded and enhanced self-renewal of hematopoietic progenitor cells (HPC), with concomitant reduction in long-term HSCs. Importantly, linoleic acid, abundant in HFD, fully recapitulates the effect of HFD on the self-renewal of PML::RARA HPCs through activation of peroxisome proliferator-activated receptor delta, a central regulator of fatty acid metabolism. Our findings inform dietary/pharmacologic interventions to counteract obesity-associated cancers and suggest that nongenetic factors play a key role., Prevention Relevance: Our work informs interventions aimed at counteracting the cancer-promoting effect of obesity. On the basis of our study, individuals with a history of chronic obesity may still significantly reduce their risk by switching to a healthier lifestyle, a concept supported by evidence in solid tumors but not yet in hematologic malignancies. See related Spotlight, p. 47., (©2023 American Association for Cancer Research.)

A maternal vaccine to protect neonates against Group B Streptococcus invasive infection is an unmet medical need. Such a vaccine should ideally be offered during the third trimester of pregnancy and induce strong immune responses after a single dose to maximize the time for placental transfer of protective antibodies. A key target antigen is the capsular polysaccharide, an anti-phagocytic virulence factor that elicits protective antibodies when conjugated to carrier proteins. The most prevalent polysaccharide serotypes conjugated to tetanus or diphtheria toxoids have been tested in humans as monovalent and multivalent formulations, showing excellent safety profiles and immunogenicity. However, responses were suboptimal in unprimed individuals after a single shot, the ideal schedule for vaccination during the third trimester of pregnancy. In the present study, we obtained and optimized self-assembling virus-like particles conjugated to Group B Streptococcus capsular polysaccharides. The resulting glyco-nanoparticles elicited strong immune responses in mice already after one immunization, providing pre-clinical proof of concept for a single-dose vaccine., (© 2023. Springer Nature Limited.)

The ability of Neisseria meningitidis Outer Membrane Vesicles (OMV) to induce protective responses in humans is well established and mainly attributed to Porin A (PorA). However, the contribution of additional protein antigens to protection remains to be elucidated. In this study we dissected the immunogenicity of antigens originating from the OMV component of the 4CMenB vaccine in mice and humans. We collected functional data on a panel of strains for which bactericidal responses to 4CMenB in infants was attributable to the OMV component and evaluated the role of 30 OMV-specific protein antigens in cross-coverage. By using tailor-made protein microarrays, the immunosignature of OMV antigens was determined. Three of these proteins, OpcA, NspA, and PorB, triggered mouse antibodies that were bactericidal against several N. meningitidis strains. Finally, by genetic deletion and/or serum depletion studies, we demonstrated the ability of OpcA and PorB to induce functional immune responses in infant sera after vaccination. In conclusion, while confirming the role of PorA in eliciting protective immunity, we identified two OMV antigens playing a key role in protection of infants vaccinated with the 4CMenB vaccine against different N. meningitidis serogroup B strains., (© 2023. The Author(s).)

Predictions of vaccine efficacy against Neisseria meningitidis serogroup B (NmB) disease are hindered by antigenic variability, limiting the representativeness of individual NmB isolates. A qualitative human serum bactericidal assay using endogenous complements of individual subjects (enc-hSBA) enables large panels of NmB isolates to be tested. A 110-isolate panel was randomly selected from 442 invasive NmB isolates from United States cases reported to the Centers for Disease Control (CDC) from 2000 to 2008. Typing analyses confirmed the 110-isolate panel is representative of the 442 isolates. The genetic features of the 110-isolate panel were compared against over 4,200 invasive NmB isolates collected from 2000 to 2018 in the United States, Australia, Canada, and nine European countries. Clonal complexes in the 110-isolate panel are also present in each geographical region; cumulative percentages show that these account for around 81% of the clonal complexes found in NmB isolates in other panels. For the antigens (fHbp, NHBA, PorA1.4, NadA) included in the currently licensed meningococcal serogroup B (MenB) vaccines, specifically considering the presence of at least one antigen with a matched genotype, the 110-isolate panel represents approximately 89% of the NmB isolates circulating worldwide, ranging from 87% for the European isolates to 95% and 97% for NmB isolates in the United States and Australia, respectively. The 110-isolate panel includes the most prevalent clonal complexes and genetic variants of MenB vaccine antigens found in a multinational collection of invasive NmB isolates. This panel is useful for assessing the efficacy of MenB vaccines in clinical trials worldwide. IMPORTANCE Neisseria meningitidis serogroup B (NmB) is a major cause of invasive meningococcal disease (IMD). Predicting the effectiveness of vaccines against NmB is difficult because NmB is an uncommon disease and because antigens targeted by meningococcal serogroup B (MenB) vaccines have highly variable genetic features and expression levels. Therefore, a large number of NmB isolates from different regions would need to be tested to comprehensively assess vaccine effectiveness. We examined a panel of 110 isolates obtained from NmB IMD cases in the United States and compared the genetic features of this panel with those of panels from different countries around the world. We found the 110-isolate panel included the most common clonal complexes and genetic variants of MenB vaccine antigens that exist in the global collections of invasive NmB isolates. This confirms the value of the NmB 110-isolate panel in understanding the effectiveness of MenB vaccines in clinical trials worldwide.

Recent studies have suggested that cardiac abnormalities can be detected from the electrocardiogram (ECG) using deep machine learning (DL) models. However, most DL algorithms lack interpretability, since they do not provide any justification for their decisions. In this study, we designed two new frameworks to interpret the classification results of DL algorithms trained for 12-lead ECG classification. The frameworks allow us to highlight not only the ECG samples that contributed most to the classification, but also which between the P-wave, QRS complex and T-wave, hereafter simply called 'waves', were the most relevant for the diagnosis. The frameworks were designed to be compatible with any DL model, including the ones already trained. The frameworks were tested on a selected Deep Neural Network, trained on a publicly available dataset, to automatically classify 24 cardiac abnormalities from 12-lead ECG signals. Experimental results showed that the frameworks were able to detect the most relevant ECG waves contributing to the classification. Often the network relied on portions of the ECG which are also considered by cardiologists to detect the same cardiac abnormalities, but this was not always the case. In conclusion, the proposed frameworks may unveil whether the network relies on features which are clinically significant for the detection of cardiac abnormalities from 12-lead ECG signals, thus increasing the trust in the DL models. This article is part of the theme issue 'Advanced computation in cardiovascular physiology: new challenges and opportunities'.

Staphylococcus aureus is a common human commensal and the leading cause of diverse infections. To identify distinctive parameters associated with infection and colonization, we compared the immune and inflammatory responses of patients with a diagnosis of invasive S. aureus disease to healthy donors. We analyzed the inflammatory responses founding a pattern of distinctive cytokines significantly higher in the patients with invasive disease. The measure of antibody levels revealed a wide antibody responsiveness from all subjects to most of the antigens, with significantly higher response for some antigens in the invasive patients compared to control. Moreover, functional antibodies against toxins distinctively associated with the invasive disease. Finally, we examined the genomic variability of isolates, showing no major differences in genetic distribution compared to a panel of representative strains. Overall, our study shows specific signatures of cytokines and functional antibodies in patients with different primary invasive diseases caused by S. aureus . These data provide insight into human responses towards invasive staphylococcal infections and are important for guiding the identification of novel preventive and therapeutic interventions against S. aureus ., Competing Interests: All authors except GG, SC, IT, and CC are employees of the GSK group of companies. FB and IM hold shares and/or restricted shares in the GSK group of companies. GG was employed by the GSK group of companies during the conduct of this study and held stock options and stocks in the GSK group of companies as part of his employee remuneration. SC was employed by the GSK group of companies during the conduct of this study. FB holds pending and issued patents on S. aureus vaccine formulations. IM and RL are listed as inventor on patents owned by the GSK group of companies. BC has non-financial support from the University of Siena outside the submitted work. The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest. The authors declare that this study received funding from Novartis Vaccines and Diagnostics Srl. The funders had the following involvement with the study: study design, conduct, analysis and data interpretation., (Copyright © 2021 Leuzzi, Bodini, Thomsen, Soldaini, Bartolini, Muzzi, Clemente, Galletti, Manetti, Giovani, Censini, Budroni, Spensieri, Borgogni, Rossi Paccani, Margarit, Bagnoli, Giudice and Creech.)

Gene expression data is commonly used in vaccine studies to characterize differences between treatment groups or sampling time points. Group-wise comparisons of the transcriptional perturbations induced by vaccination have been applied extensively for investigating the mechanisms of action of vaccines. Such approaches, however, may not be sensitive enough for detecting changes occurring within a minority of the population under investigation or in single individuals. In this study, we developed a data analysis framework to characterize individual subject response profiles in the context of repeated measure experiments, which are typical of vaccine mode of action studies. Following the definition of the methodology, this was applied to the analysis of human transcriptome responses induced by vaccination with a subunit influenza vaccine. Results highlighted a substantial heterogeneity in how different subjects respond to vaccination. Moreover, the extent of transcriptional modulation experienced by each individual subject was found to be associated with the magnitude of vaccine-specific functional antibody response, pointing to a mechanistic link between genes involved in protein production and innate antiviral response. Overall, we propose that the improved characterization of the intersubject heterogeneity, enabled by our approach, can help driving the improvement and optimization of current and next-generation vaccines., (© 2021. The Author(s).)

Aims : Bubble entropy (bEn) is an entropy metric with a limited dependence on parameters. bEn does not directly quantify the conditional entropy of the series, but it assesses the change in entropy of the ordering of portions of its samples of length m , when adding an extra element. The analytical formulation of bEn for autoregressive (AR) processes shows that, for this class of processes, the relation between the first autocorrelation coefficient and bEn changes for odd and even values of m . While this is not an issue, per se, it triggered ideas for further investigation. Methods : Using theoretical considerations on the expected values for AR processes, we examined a two-steps-ahead estimator of bEn, which considered the cost of ordering two additional samples. We first compared it with the original bEn estimator on a simulated series. Then, we tested it on real heart rate variability (HRV) data. Results : The experiments showed that both examined alternatives showed comparable discriminating power. However, for values of 10

An increase in invasive meningococcal disease (IMD) incidence was observed in Tuscany in 2015/2016, mainly due to hypervirulent clonal complex (cc) 11 strains. In a post-hoc analysis, we assessed bactericidal activity of antibodies in sera from children primed with MenACWY-CRM or MenC-CRM conjugate vaccines and receiving a MenACWY-CRM booster dose against 5 meningococcal C (MenC) strains isolated from IMD cases. Sera collected from 90 infants/toddlers who participated in a phase III, open-label study (NCT00667602) and its extension (NCT01345721) were tested by serum bactericidal activity assay with human complement (hSBA). Children were primed with either MenACWY-CRM at 6-8 and 12 months of age (group 2_MenACWY; N = 30), MenACWY-CRM (group 1_MenACWY; N = 30), or MenC-CRM at 12 months of age (group 1_MenC; N = 30); all received MenACWY-CRM booster dose at 22-45 months of age. Four tested strains (FI001-FI004) were C:P1.5-1,10-8:F3-6:ST-11 (cc11) and 1 (FI005) was C:P1.7-4,14-6:F3-9:ST-1031 (cc334). Overall, immune responses tended to be higher against Fl002-FI004 than Fl001 and Fl005. Geometric mean titers were high in group 2_MenACWY (range: 94.8 [FI005]-588.1 [FI004]) and very high post-boosting with MenACWY-CRM in all groups (176.9 [FI005]-3911.0 [FI004]). Seroresponse rates tended to be higher in group 1_MenC (33.3% [FI005]-93.3% [FI004]) than in group 1_MenACWY (16.7% [FI005]-73.3% [FI004]). Irrespective of strains tested or the identity/number of priming doses, ≥96.7% of children had hSBA titers ≥1:8 post-MenACWY-CRM booster dose. MenACWY-CRM and MenC-CRM elicited bactericidal antibodies and immunological memory against hypervirulent cc11 and cc334 MenC strains responsible for IMD outbreaks.

Neisseria meningitidis is a strictly human pathogen and is the major cause of septicemia and meningitis worldwide. Factor H binding protein (fHbp) is a meningococcal surface-exposed lipoprotein that binds the human Complement factor H allowing the bacterium to evade the host innate immune response. FHbp is also a key antigen in two vaccines against N. meningitidis serogroup B. Although the fHbp gene is present in most circulating meningococcal strains, level of fHbp expression varies among isolates and has been correlated to differences in promoter sequences upstream of the gene. Here we elucidated the sequence determinants that control fHbp expression in globally circulating strains. We analyzed the upstream fHbp intergenic region (fIR) of more than 5800 strains representative of the UK circulating isolates and we identified eleven fIR sequence alleles which represent 88% of meningococcal strains. By engineering isogenic recombinant strains where fHbp expression was under the control of each of the eleven fIR alleles, we confirmed that the fIR sequence determines a specific and distinct level of expression. Moreover, we identified the molecular basis for variation in expression through polymorphisms within key regulatory regions that are known to affect fHbp expression. We experimentally established three expression groups, high-medium-low, that correlated directly with the susceptibility to killing mediated by anti-fHbp antibodies and the ability of the meningococcal strain to survive within human serum. By using this sequence classification and information about the variant, we predicted fHbp expression in the panel of UK strains and we observed that strains with higher expressing fIR alleles are more likely associated with invasive disease. Overall, our findings can contribute to understand and predict vaccine coverage mediated by fHbp as well as to shed light on the role of this virulence factor in determining an invasive phenotype., Competing Interests: I have read the journal’s policy and the authors of this manuscript have the following competing interests: MS and MB were employees of Randstad Italia S.p.A, working as contractors for GSK at the time the study was conducted, and are now employees of the GSK group of companies. LF, AB, AM and ID are employees of the GSK group of companies; MDC was an employee of the GSK group of companies, now at Université Côte d’Azur. ID reports ownership of GSK stocks. ID is listed as inventor on patents on vaccine candidates owned by the GSK group of companies. At the time of the study MS and TB were recipients of a GSK fellowship from the PhD program of the University of Bologna. VS declares no conflicts of interest.

Invasive meningococcal disease (IMD) caused by Neisseria meningitidis is a significant cause of morbidity and mortality worldwide. In Finland, the incidence rate of IMD is low, with meningococcal serogroup B (MenB) accounting for around one-third of IMD cases annually. The aim of this study was to investigate the genetic variability of invasive MenB isolates collected in Finland between 2010 and 2017 ( n  = 81), including the genes encoding the 4-component MenB vaccine (4CMenB; Bexsero; GSK) antigens and their promoters, and to evaluate the 4CMenB potential coverage. Whole-genome sequencing was performed. The meningococcal antigen typing system (MATS) was used to characterize MenB isolates and predict the potential coverage of 4CMenB. MATS was complemented by genetic MATS (gMATS) through association of antigen genotyping and phenotypic MATS results. Multilocus sequence typing revealed predominance of the ST-41/44 clonal complex among which sequence type (ST)-303 was the most common and was predicted to be covered by 4CMenB. Of the 4 major vaccine antigens, the factor H-binding protein variant 1, neisserial heparin binding antigen peptide 2, and the PorA P1.4 antigen were predominant, whereas Neisseria adhesin A was present in only 4% of the 81 isolates. MATS and gMATS 4CMenB strain coverage predictions were 78% and 86%, respectively, in a subpanel of 60 isolates collected during 2010 to 2014, with a gMATS prediction of 84% for all 81 isolates. The results suggest that 4CMenB could reduce the burden of IMD in Finland and that gMATS could be applied to monitor vaccine strain coverage and predict vaccine effectiveness. IMPORTANCE 4CMenB is a 4-component vaccine used against invasive meningococcal disease (IMD) caused by Neisseria meningitidis serogroup B (MenB). We investigated the genetic variability of MenB in Finland and evaluated 4CMenB strain coverage by 2 different methods: MATS (meningococcal antigen typing system) and gMATS (genetic MATS). In a set of MenB isolates, 78% (MATS) and 86% (gMATS) were predicted as covered by 4CMenB, suggesting that use of 4CMenB would help reduce IMD incidence in Finland. MATS has been used in 13 countries worldwide, generating information on phenotypic characteristics that could infer protection by 4CMenB. Based on these data and genetic information, gMATS coverage predictions can be made. gMATS predicts coverage consistent with MATS for about 94% of tested strains. Unlike MATS, gMATS does not require live isolates, thus allowing the analysis also of noncultivable strains, making the coverage predictions more accurate. Therefore, gMATS can replace MATS to assess 4CMenB coverage, including in regions with no prior MATS data., (Copyright © 2020 Bodini et al.)

Background: Multi-locus sequence typing (MLST) is a standard typing technique used to associate a sequence type (ST) to a bacterial isolate. When the output of whole genome sequencing (WGS) of a sample is available the ST can be assigned directly processing the read-set. Current approaches employ reads mapping (SRST2) against the MLST loci, k-mer distribution (stringMLST), selective assembly (GRAbB) or whole genome assembly (BIGSdb) followed by BLASTn sequence query. Here we present STRAIN (ST Reduced Assembly IdentificatioN), an R package that implements a hybrid strategy between assembly and mapping of the reads to assign the ST to an isolate starting from its read-sets., Results: Analysis of 540 publicly accessible Illumina read sets showed STRAIN to be more accurate at correct allele assignment and new alleles identification compared to SRTS2, stringMLST and GRAbB. STRAIN assigned correctly 3666 out of 3780 alleles (capability to identify correct alleles 97%) and, when presented with samples containing new alleles, identified them in 3730 out of 3780 STs (capability to identify new alleles 98.7%) of the cases. On the same dataset the other tested tools achieved lower capability to identify correct alleles (from 28.5 to 96.9%) and lower capability to identify new alleles (from 1.1 to 97.1%)., Conclusions: STRAIN is a new accurate method to assign the alleles and ST to an isolate by processing the raw reads output of WGS. STRAIN is also able to retrieve new allele sequences if present. Capability to identify correct and new STs/alleles, evaluated on a benchmark dataset, are higher than other existing methods. STRAIN is designed for single allele typing as well as MLST. Its implementation in R makes allele and ST assignment simple, direct and prompt to be integrated in wider pipeline of downstream bioinformatics analyses.

Hardware accelerators based on two-terminal non-volatile memories (NVMs) can potentially provide competitive speed and accuracy for the training of fully connected deep neural networks (FC-DNNs), with respect to GPUs and other digital accelerators. We recently proposed [S. Ambrogio et al., Nature, 2018] novel neuromorphic crossbar arrays, consisting of a pair of phase-change memory (PCM) devices combined with a pair of 3-Transistor 1-Capacitor (3T1C) circuit elements, so that each weight was implemented using multiple conductances of varying significance, and then showed that this weight element can train FC-DNNs to software-equivalent accuracies. Unfortunately, however, real arrays of emerging NVMs such as PCM typically include some failed devices (e.g., <100% yield), either due to fabrication issues or early endurance failures, which can degrade DNN training accuracy. This paper explores the impact of device failures, NVM conductances that may contribute read current but which cannot be programmed, on DNN training and test accuracy. Results show that "stuck-on" and "dead" devices, exhibiting high and low read conductances, respectively, do in fact degrade accuracy performance to some degree. We find that the presence of the CMOS-based and thus highly-reliable 3T1C devices greatly increase system robustness. After studying the inherent mechanisms, we study the dependence of DNN accuracy on the number of functional weights, the number of neurons in the hidden layer, and the number and type of damaged devices. Finally, we describe conditions under which making the network larger or adjusting the network hyperparameters can still improve the network accuracy, even in the presence of failed devices.

Background: The Meningococcal Antigen Typing System (MATS) was developed to identify meningococcus group B strains with a high likelihood of being covered by the 4CMenB vaccine, but is limited by the requirement for viable isolates from culture-confirmed cases. We examined if antigen genotyping could complement MATS in predicting strain coverage by the 4CMenB vaccine., Methods: From a panel of 3912 MATS-typed invasive meningococcal disease isolates collected in England and Wales in 2007-2008, 2014-2015 and 2015-2016, and in 16 other countries in 2000-2015, 3481 isolates were also characterized by antigen genotyping. Individual associations between antigen genotypes and MATS coverage for each 4CMenB component were used to define a genetic MATS (gMATS). gMATS estimates were compared with England and Wales human complement serum bactericidal assay (hSBA) data and vaccine effectiveness (VE) data from England., Results: Overall, 81% of the strain panel had genetically predictable MATS coverage, with 92% accuracy and highly concordant results across national panels (Lin's accuracy coefficient, 0.98; root-mean-square deviation, 6%). England and Wales strain coverage estimates were 72-73% by genotyping (66-73% by MATS), underestimating hSBA values after four vaccine doses (88%) and VE after two doses (83%). The gMATS predicted strain coverage in other countries was 58-88%., Conclusions: gMATS can replace MATS in predicting 4CMenB strain coverage in four out of five cases, without requiring a cultivable isolate, and is open to further improvement. Both methods underestimated VE in England. Strain coverage predictions in other countries matched or exceeded England and Wales estimates., (Copyright © 2019 GlaxoSmithKline Biologicals SA. Published by Elsevier Ltd.. All rights reserved.)

Neural-network training can be slow and energy intensive, owing to the need to transfer the weight data for the network between conventional digital memory chips and processor chips. Analogue non-volatile memory can accelerate the neural-network training algorithm known as backpropagation by performing parallelized multiply-accumulate operations in the analogue domain at the location of the weight data. However, the classification accuracies of such in situ training using non-volatile-memory hardware have generally been less than those of software-based training, owing to insufficient dynamic range and excessive weight-update asymmetry. Here we demonstrate mixed hardware-software neural-network implementations that involve up to 204,900 synapses and that combine long-term storage in phase-change memory, near-linear updates of volatile capacitors and weight-data transfer with 'polarity inversion' to cancel out inherent device-to-device variations. We achieve generalization accuracies (on previously unseen data) equivalent to those of software-based training on various commonly used machine-learning test datasets (MNIST, MNIST-backrand, CIFAR-10 and CIFAR-100). The computational energy efficiency of 28,065 billion operations per second per watt and throughput per area of 3.6 trillion operations per second per square millimetre that we calculate for our implementation exceed those of today's graphical processing units by two orders of magnitude. This work provides a path towards hardware accelerators that are both fast and energy efficient, particularly on fully connected neural-network layers.

Background: The landscape of cancer-predisposing genes has been extensively investigated in the last 30 years with various methodologies ranging from candidate gene to genome-wide association studies. However, sequencing data are still poorly exploited in cancer predisposition studies due to the lack of statistical power when comparing millions of variants at once., Method: To overcome these power limitations, we propose a knowledge-based framework founded on the characteristics of known cancer-predisposing variants and genes. Under our framework, we took advantage of a combination of previously generated datasets of sequencing experiments to identify novel breast cancer-predisposing variants, comparing the normal genomes of 673 breast cancer patients of European origin against 27,173 controls matched by ethnicity., Results: We detected several expected variants on known breast cancer-predisposing genes, like BRCA1 and BRCA2, and 11 variants on genes associated with other cancer types, like RET and AKT1. Furthermore, we detected 183 variants that overlap with somatic mutations in cancer and 41 variants associated with 38 possible loss-of-function genes, including PIK3CB and KMT2C. Finally, we found a set of 19 variants that are potentially pathogenic, negatively correlate with age at onset, and have never been associated with breast cancer., Conclusions: In this study, we demonstrate the usefulness of a genomic-driven approach nested in a classic case-control study to prioritize cancer-predisposing variants. In addition, we provide a resource containing variants that may affect susceptibility to breast cancer.