Search

Background: Cancer is the leading cause of disease-related mortality in children. Causes of leukemia, the most common form, are largely unknown. Growing evidence points to an origin in-utero, when global redistribution of DNA methylation occurs driving tissue differentiation. Methods: Epigenome-wide DNA methylation was profiled in surrogate (blood) and target (bone marrow) tissues at birth, diagnosis, remission and relapse of pediatric pre-B acute lymphoblastic leukemia (pre-B ALL) patients. Double-blinded analyses was performed between prospective cohorts extending from birth to diagnosis and retrospective studies backtracking from clinical disease to birth. Validation was carried out using independent technologies and populations. Results: The imprinted and immuno-modulating VTRNA2-1 was hypermethylated (FDR<0.05) at birth in nested cases relative to controls in all tested populations (totaling 317 cases and 483 controls), including European and Hispanic ancestries. VTRNA2-1 methylation was stable over follow-up years after birth and across surrogate, target and other tissues (n=5,023 tissues; 30 types). When profiled in leukemic tissues from two clinical cohorts (totaling 644 cases), VTRNA2-1 methylation exhibited higher levels at diagnosis relative to controls, it reset back to normal levels at remission, and then re-increased to above control levels at relapse. Hypermethylation was significantly associated with worse pre-B ALL patient survival and with reduced VTRNA2-1 expression (n=2,294 tissues; 26 types), supporting a functional and translational role for VTRNA2-1 methylation. Conclusion: This study provides proof-of-concept to detect at birth epigenetic precursors of pediatric pre-B ALL. These alterations were reproducible with different technologies, in three continents and in two ethnicities, and can offer biomarkers for early detection and prognosis as well as actionable targets for therapy.

Higher birth order is associated with altered risk of many disease states. Changes in placentation and exposures to in utero growth factors with successive pregnancies may impact later life disease risk via persistent DNA methylation alterations. We investigated birth order with Illumina DNA methylation array data in each of 16 birth cohorts (8164 newborns) with European, African, and Latino ancestries from the Pregnancy and Childhood Epigenetics Consortium. Meta-analyzed data demonstrated systematic DNA methylation variation in 341 CpGs (FDR adjusted P < 0.05) and 1107 regions. Forty CpGs were located within known quantitative trait loci for gene expression traits in blood, and trait enrichment analysis suggested a strong association with immune-related, transcriptional control, and blood pressure regulation phenotypes. Decreasing fertility rates worldwide with the concomitant increased proportion of first-born children highlights a potential reflection of birth order-related epigenomic states on changing disease incidence trends.

Joint consideration of scheduling and adaptive parallelism offers great opportunities for improving the training efficiency of large models on heterogeneous GPU clusters. However, integrating adaptive parallelism into a cluster scheduler expands the cluster scheduling space. The new space is the product of the original scheduling space and the parallelism exploration space of adaptive parallelism (also a product of pipeline, data, and tensor parallelism). The exponentially enlarged scheduling space and ever-changing optimal parallelism plan from adaptive parallelism together result in the contradiction between low-overhead and accurate performance data acquisition for efficient cluster scheduling. This paper presents Crius, a training system for efficiently scheduling multiple large models with adaptive parallelism in a heterogeneous cluster. Crius proposes a novel scheduling granularity called Cell. It represents a job with deterministic resources and pipeline stages. The exploration space of Cell is shrunk to the product of only data and tensor parallelism, thus exposing the potential for accurate and low-overhead performance estimation. Crius then accurately estimates Cells and efficiently schedules training jobs. When a Cell is selected as a scheduling choice, its represented job runs with the optimal parallelism plan explored. Experimental results show that Crius reduces job completion time by up to 48.9% and schedules large models with up to 1.49x cluster throughput improvement.

BackgroundChildhood glioblastoma multiforme (GBM) is a highly aggressive disease with low survival, and its etiology, especially concerning germline genetic risk, is poorly understood. Mitochondria play a key role in putative tumorigenic processes relating to cellular oxidative metabolism, and mitochondrial DNA variants were not previously assessed for association with pediatric brain tumor risk.MethodsWe conducted an analysis of 675 mitochondrial DNA variants in 90 childhood GBM cases and 2789 controls to identify enrichment of mitochondrial variant associated with GBM risk. We also performed this analysis for other glioma subtypes including pilocytic astrocytoma. Nuclear-encoded mitochondrial gene variants were also analyzed.ResultsWe identified m1555 A>G was significantly associated with GBM risk (adjusted OR 29.30, 95% CI 5.25-163.4, P-value 9.5 X 10-4). No association was detected for other subtypes. Haplotype analysis further supported the independent risk contributed by m1555 G>A, instead of a haplogroup joint effect. Nuclear-encoded mitochondrial gene variants identified significant associations in European (rs62036057 in WWOX, adjusted OR = 2.99, 95% CI 1.88-4.75, P-value = 3.42 X 10-6) and Hispanic (rs111709726 in EFHD1, adjusted OR = 3.57, 95% CI 1.99-6.40, P-value = 1.41 X 10-6) populations in ethnicity-stratified analyses.ConclusionWe report for the first time a potential role played by a functional mitochondrial ribosomal RNA variant in childhood GBM risk, and a potential role for both mitochondrial and nuclear-mitochondrial DNA polymorphisms in GBM tumorigenesis. These data implicate cellular oxidative metabolic capacity as a contributor to the etiology of pediatric glioblastoma.

BackgroundChildhood glioblastoma multiforme (GBM) is a highly aggressive disease with low survival, and its etiology, especially concerning germline genetic risk, is poorly understood. Mitochondria play a key role in putative tumorigenic processes relating to cellular oxidative metabolism, and mitochondrial DNA variants were not previously assessed for association with pediatric brain tumor risk.MethodsWe conducted an analysis of 675 mitochondrial DNA variants in 90 childhood GBM cases and 2789 controls to identify enrichment of mitochondrial variant associated with GBM risk. We also performed this analysis for other glioma subtypes including pilocytic astrocytoma. Nuclear-encoded mitochondrial gene variants were also analyzed.ResultsWe identified m1555 A>G was significantly associated with GBM risk (adjusted OR 29.30, 95% CI 5.25-163.4, P-value 9.5 X 10-4). No association was detected for other subtypes. Haplotype analysis further supported the independent risk contributed by m1555 G>A, instead of a haplogroup joint effect. Nuclear-encoded mitochondrial gene variants identified significant associations in European (rs62036057 in WWOX, adjusted OR = 2.99, 95% CI 1.88-4.75, P-value = 3.42 X 10-6) and Hispanic (rs111709726 in EFHD1, adjusted OR = 3.57, 95% CI 1.99-6.40, P-value = 1.41 X 10-6) populations in ethnicity-stratified analyses.ConclusionWe report for the first time a potential role played by a functional mitochondrial ribosomal RNA variant in childhood GBM risk, and a potential role for both mitochondrial and nuclear-mitochondrial DNA polymorphisms in GBM tumorigenesis. These data implicate cellular oxidative metabolic capacity as a contributor to the etiology of pediatric glioblastoma.

BackgroundPilocytic astrocytoma (PA) is the most common pediatric brain tumor. PA has at least a 50% higher incidence in populations of European ancestry compared to other ancestral groups, which may be due in part to genetic differences.MethodsWe first compared the global proportions of European, African, and Amerindian ancestries in 301 PA cases and 1185 controls of self-identified Latino ethnicity from the California Biobank. We then conducted admixture mapping analysis to assess PA risk with local ancestry.ResultsWe found PA cases had a significantly higher proportion of global European ancestry than controls (case median = 0.55, control median = 0.51, P value = 3.5x10-3). Admixture mapping identified 13 SNPs in the 6q14.3 region (SNX14) contributing to risk, as well as three other peaks approaching significance on chromosomes 7, 10 and 13. Downstream fine mapping in these regions revealed several SNPs potentially contributing to childhood PA risk.ConclusionsThere is a significant difference in genomic ancestry associated with Latino PA risk and several genomic loci potentially mediating this risk.

Background Although recent sequencing studies have revealed that 10% of childhood gliomas are caused by rare germline mutations, the role of common variants is undetermined and no genome-wide significant risk loci for pediatric central nervous system tumors have been identified to date. Methods Meta-analysis of 3 population-based genome-wide association studies comprising 4069 children with glioma and 8778 controls of multiple genetic ancestries. Replication was performed in a separate case–control cohort. Quantitative trait loci analyses and a transcriptome-wide association study were conducted to assess possible links with brain tissue expression across 18 628 genes. Results Common variants in CDKN2B-AS1 at 9p21.3 were significantly associated with astrocytoma, the most common subtype of glioma in children (rs573687, P-value of 6.974e-10, OR 1.273, 95% CI 1.179–1.374). The association was driven by low-grade astrocytoma (P-value of 3.815e-9) and exhibited unidirectional effects across all 6 genetic ancestries. For glioma overall, the association approached genome-wide significance (rs3731239, P-value of 5.411e-8), while no significant association was observed for high-grade tumors. Predicted decreased brain tissue expression of CDKN2B was significantly associated with astrocytoma (P-value of 8.090e-8). Conclusions In this population-based genome-wide association study meta-analysis, we identify and replicate 9p21.3 (CDKN2B-AS1) as a risk locus for childhood astrocytoma, thereby establishing the first genome-wide significant evidence of common variant predisposition in pediatric neuro-oncology. We furthermore provide a functional basis for the association by showing a possible link to decreased brain tissue CDKN2B expression and substantiate that genetic susceptibility differs between low- and high-grade astrocytoma., Background: Although recent sequencing studies have revealed that 10% of childhood gliomas are caused by rare germline mutations, the role of common variants is undetermined and no genome-wide significant risk loci for pediatric central nervous system tumors have been identified to date. Methods: Meta-analysis of 3 population-based genome-wide association studies comprising 4069 children with glioma and 8778 controls of multiple genetic ancestries. Replication was performed in a separate case-control cohort. Quantitative trait loci analyses and a transcriptome-wide association study were conducted to assess possible links with brain tissue expression across 18 628 genes. Results: Common variants in CDKN2B-AS1 at 9p21.3 were significantly associated with astrocytoma, the most common subtype of glioma in children (rs573687, P-value of 6.974e-10, OR 1.273, 95% CI 1.179-1.374). The association was driven by low-grade astrocytoma (P-value of 3.815e-9) and exhibited unidirectional effects across all 6 genetic ancestries. For glioma overall, the association approached genome-wide significance (rs3731239, P-value of 5.411e-8), while no significant association was observed for high-grade tumors. Predicted decreased brain tissue expression of CDKN2B was significantly associated with astrocytoma (P-value of 8.090e-8). Conclusions: In this population-based genome-wide association study meta-analysis, we identify and replicate 9p21.3 (CDKN2B-AS1) as a risk locus for childhood astrocytoma, thereby establishing the first genome-wide significant evidence of common variant predisposition in pediatric neuro-oncology. We furthermore provide a functional basis for the association by showing a possible link to decreased brain tissue CDKN2B expression and substantiate that genetic susceptibility differs between low- and high-grade astrocytoma.

Background Although recent sequencing studies have revealed that 10% of childhood gliomas are caused by rare germline mutations, the role of common variants is undetermined and no genome-wide significant risk loci for pediatric central nervous system tumors have been identified to date. Methods Meta-analysis of 3 population-based genome-wide association studies comprising 4069 children with glioma and 8778 controls of multiple genetic ancestries. Replication was performed in a separate case–control cohort. Quantitative trait loci analyses and a transcriptome-wide association study were conducted to assess possible links with brain tissue expression across 18 628 genes. Results Common variants in CDKN2B-AS1 at 9p21.3 were significantly associated with astrocytoma, the most common subtype of glioma in children (rs573687, P-value of 6.974e-10, OR 1.273, 95% CI 1.179–1.374). The association was driven by low-grade astrocytoma (P-value of 3.815e-9) and exhibited unidirectional effects across all 6 genetic ancestries. For glioma overall, the association approached genome-wide significance (rs3731239, P-value of 5.411e-8), while no significant association was observed for high-grade tumors. Predicted decreased brain tissue expression of CDKN2B was significantly associated with astrocytoma (P-value of 8.090e-8). Conclusions In this population-based genome-wide association study meta-analysis, we identify and replicate 9p21.3 (CDKN2B-AS1) as a risk locus for childhood astrocytoma, thereby establishing the first genome-wide significant evidence of common variant predisposition in pediatric neuro-oncology. We furthermore provide a functional basis for the association by showing a possible link to decreased brain tissue CDKN2B expression and substantiate that genetic susceptibility differs between low- and high-grade astrocytoma., Background: Although recent sequencing studies have revealed that 10% of childhood gliomas are caused by rare germline mutations, the role of common variants is undetermined and no genome-wide significant risk loci for pediatric central nervous system tumors have been identified to date. Methods: Meta-analysis of 3 population-based genome-wide association studies comprising 4069 children with glioma and 8778 controls of multiple genetic ancestries. Replication was performed in a separate case-control cohort. Quantitative trait loci analyses and a transcriptome-wide association study were conducted to assess possible links with brain tissue expression across 18 628 genes. Results: Common variants in CDKN2B-AS1 at 9p21.3 were significantly associated with astrocytoma, the most common subtype of glioma in children (rs573687, P-value of 6.974e-10, OR 1.273, 95% CI 1.179-1.374). The association was driven by low-grade astrocytoma (P-value of 3.815e-9) and exhibited unidirectional effects across all 6 genetic ancestries. For glioma overall, the association approached genome-wide significance (rs3731239, P-value of 5.411e-8), while no significant association was observed for high-grade tumors. Predicted decreased brain tissue expression of CDKN2B was significantly associated with astrocytoma (P-value of 8.090e-8). Conclusions: In this population-based genome-wide association study meta-analysis, we identify and replicate 9p21.3 (CDKN2B-AS1) as a risk locus for childhood astrocytoma, thereby establishing the first genome-wide significant evidence of common variant predisposition in pediatric neuro-oncology. We furthermore provide a functional basis for the association by showing a possible link to decreased brain tissue CDKN2B expression and substantiate that genetic susceptibility differs between low- and high-grade astrocytoma.

BACKGROUND: Although recent sequencing studies have revealed that 10% of childhood gliomas are caused by rare germline mutations, the role of common variants is undetermined and no genome-wide significant risk loci for pediatric central nervous system tumors have been identified to date. METHODS: Meta-analysis of 3 population-based genome-wide association studies comprising 4069 children with glioma and 8778 controls of multiple genetic ancestries. Replication was performed in a separate case-control cohort. Quantitative trait loci analyses and a transcriptome-wide association study were conducted to assess possible links with brain tissue expression across 18 628 genes. RESULTS: Common variants in CDKN2B-AS1 at 9p21.3 were significantly associated with astrocytoma, the most common subtype of glioma in children (rs573687, P-value of 6.974e-10, OR 1.273, 95% CI 1.179-1.374). The association was driven by low-grade astrocytoma (P-value of 3.815e-9) and exhibited unidirectional effects across all 6 genetic ancestries. For glioma overall, the association approached genome-wide significance (rs3731239, P-value of 5.411e-8), while no significant association was observed for high-grade tumors. Predicted decreased brain tissue expression of CDKN2B was significantly associated with astrocytoma (P-value of 8.090e-8). CONCLUSIONS: In this population-based genome-wide association study meta-analysis, we identify and replicate 9p21.3 (CDKN2B-AS1) as a risk locus for childhood astrocytoma, thereby establishing the first genome-wide significant evidence of common variant predisposition in pediatric neuro-oncology. We furthermore provide a functional basis for the association by showing a possible link to decreased brain tissue CDKN2B expression and substantiate that genetic susceptibility differs between low- and high-grade astrocytoma.

BACKGROUND: Although recent sequencing studies have revealed that 10% of childhood gliomas are caused by rare germline mutations, the role of common variants is undetermined and no genome-wide significant risk loci for pediatric central nervous system tumors have been identified to date. METHODS: Meta-analysis of 3 population-based genome-wide association studies comprising 4069 children with glioma and 8778 controls of multiple genetic ancestries. Replication was performed in a separate case-control cohort. Quantitative trait loci analyses and a transcriptome-wide association study were conducted to assess possible links with brain tissue expression across 18 628 genes. RESULTS: Common variants in CDKN2B-AS1 at 9p21.3 were significantly associated with astrocytoma, the most common subtype of glioma in children (rs573687, P-value of 6.974e-10, OR 1.273, 95% CI 1.179-1.374). The association was driven by low-grade astrocytoma (P-value of 3.815e-9) and exhibited unidirectional effects across all 6 genetic ancestries. For glioma overall, the association approached genome-wide significance (rs3731239, P-value of 5.411e-8), while no significant association was observed for high-grade tumors. Predicted decreased brain tissue expression of CDKN2B was significantly associated with astrocytoma (P-value of 8.090e-8). CONCLUSIONS: In this population-based genome-wide association study meta-analysis, we identify and replicate 9p21.3 (CDKN2B-AS1) as a risk locus for childhood astrocytoma, thereby establishing the first genome-wide significant evidence of common variant predisposition in pediatric neuro-oncology. We furthermore provide a functional basis for the association by showing a possible link to decreased brain tissue CDKN2B expression and substantiate that genetic susceptibility differs between low- and high-grade astrocytoma.

BackgroundGenetic variants can modulate phenotypic outcomes via epigenetic intermediates, for example at methylation quantitative trait loci (mQTL). We present the first large-scale assessment of mQTL at human genomic regions selected for interindividual variation in CpG methylation, which we call correlated regions of systemic interindividual variation (CoRSIVs). These can be assayed in blood DNA and do not reflect interindividual variation in cellular composition.ResultsWe use target-capture bisulfite sequencing to assess DNA methylation at 4086 CoRSIVs in multiple tissues from each of 188 donors in the NIH Gene-Tissue Expression (GTEx) program. At CoRSIVs, DNA methylation in peripheral blood correlates with methylation and gene expression in internal organs. We also discover unprecedented mQTL at these regions. Genetic influences on CoRSIV methylation are extremely strong (median R2=0.76), cumulatively comprising over 70-fold more human mQTL than detected in the most powerful previous study. Moreover, mQTL beta coefficients at CoRSIVs are highly skewed (i.e., the major allele predicts higher methylation). Both surprising findings are independently validated in a cohort of 47 non-GTEx individuals. Genomic regions flanking CoRSIVs show long-range enrichments for LINE-1 and LTR transposable elements; the skewed beta coefficients may therefore reflect evolutionary selection of genetic variants that promote their methylation and silencing. Analyses of GWAS summary statistics show that mQTL polymorphisms at CoRSIVs are associated with metabolic and other classes of disease.ConclusionsA focus on systemic interindividual epigenetic variants, clearly enhanced in mQTL content, should likewise benefit studies attempting to link human epigenetic variation to the risk of disease.

BACKGROUND: Although recent sequencing studies have revealed that 10% of childhood gliomas are caused by rare germline mutations, the role of common variants is undetermined and no genome-wide significant risk loci for pediatric central nervous system tumors have been identified to date. METHODS: Meta-analysis of 3 population-based genome-wide association studies comprising 4069 children with glioma and 8778 controls of multiple genetic ancestries. Replication was performed in a separate case-control cohort. Quantitative trait loci analyses and a transcriptome-wide association study were conducted to assess possible links with brain tissue expression across 18 628 genes. RESULTS: Common variants in CDKN2B-AS1 at 9p21.3 were significantly associated with astrocytoma, the most common subtype of glioma in children (rs573687, P-value of 6.974e-10, OR 1.273, 95% CI 1.179-1.374). The association was driven by low-grade astrocytoma (P-value of 3.815e-9) and exhibited unidirectional effects across all 6 genetic ancestries. For glioma overall, the association approached genome-wide significance (rs3731239, P-value of 5.411e-8), while no significant association was observed for high-grade tumors. Predicted decreased brain tissue expression of CDKN2B was significantly associated with astrocytoma (P-value of 8.090e-8). CONCLUSIONS: In this population-based genome-wide association study meta-analysis, we identify and replicate 9p21.3 (CDKN2B-AS1) as a risk locus for childhood astrocytoma, thereby establishing the first genome-wide significant evidence of common variant predisposition in pediatric neuro-oncology. We furthermore provide a functional basis for the association by showing a possible link to decreased brain tissue CDKN2B expression and substantiate that genetic susceptibility differs between low- and high-grade astrocytoma.

BACKGROUND: Although recent sequencing studies have revealed that 10% of childhood gliomas are caused by rare germline mutations, the role of common variants is undetermined and no genome-wide significant risk loci for pediatric central nervous system tumors have been identified to date. METHODS: Meta-analysis of 3 population-based genome-wide association studies comprising 4069 children with glioma and 8778 controls of multiple genetic ancestries. Replication was performed in a separate case-control cohort. Quantitative trait loci analyses and a transcriptome-wide association study were conducted to assess possible links with brain tissue expression across 18 628 genes. RESULTS: Common variants in CDKN2B-AS1 at 9p21.3 were significantly associated with astrocytoma, the most common subtype of glioma in children (rs573687, P-value of 6.974e-10, OR 1.273, 95% CI 1.179-1.374). The association was driven by low-grade astrocytoma (P-value of 3.815e-9) and exhibited unidirectional effects across all 6 genetic ancestries. For glioma overall, the association approached genome-wide significance (rs3731239, P-value of 5.411e-8), while no significant association was observed for high-grade tumors. Predicted decreased brain tissue expression of CDKN2B was significantly associated with astrocytoma (P-value of 8.090e-8). CONCLUSIONS: In this population-based genome-wide association study meta-analysis, we identify and replicate 9p21.3 (CDKN2B-AS1) as a risk locus for childhood astrocytoma, thereby establishing the first genome-wide significant evidence of common variant predisposition in pediatric neuro-oncology. We furthermore provide a functional basis for the association by showing a possible link to decreased brain tissue CDKN2B expression and substantiate that genetic susceptibility differs between low- and high-grade astrocytoma.

BackgroundAssociations between maternal tobacco exposure during pregnancy and childhood acute lymphoblastic leukemia (ALL) have yielded mixed results. This may be due to biases in self-reported smoking or other differences in individual level risk factors. We utilized a biological marker of maternal tobacco exposure to evaluate the association between maternal tobacco exposure during pregnancy, genetics, and subsequent childhood ALL risk in two large population-based studies of childhood ALL in California.MethodsMaternal exposure to tobacco smoke was assessed with a validated methylation marker (cg05575921) of the aryl hydrocarbon receptor repressor (AHRR) gene in newborn dried blood spots. We adjusted for sex, birthweight, gestational age, mode of delivery, year of birth, AHRR quantitative trait locus (mQTL) rs77111113, and a polygenetic risk score for childhood ALL. We additionally adjusted for principal components in a gene-environment interaction testing method that incorporates gene-only and environment-only effects along with interactions.ResultsAHRR hypomethylation overall was not associated with childhood ALL. In gene-environment interaction testing, several genetic variants displayed significant interaction with AHRR hypomethylation and childhood ALL.ConclusionsOur results suggest novel candidates in PTPRK and DPP6 may play a role in tobacco-related leukemogenesis. Further research is necessary to better understand the effects of tobacco and these variants on childhood ALL risk.ImpactDespite lack of an overall "main effect," tobacco exposure during pregnancy impacts childhood ALL risk depending on specific genetic variants.

BACKGROUND: Congenital heart defects (CHDs) affect approximately half of individuals with Down syndrome (DS), but the molecular reasons for incomplete penetrance are unknown. Previous studies have largely focused on identifying genetic risk factors associated with CHDs in individuals with DS, but comprehensive studies of the contribution of epigenetic marks are lacking. We aimed to identify and characterize DNA methylation differences from newborn dried blood spots (NDBS) of DS individuals with major CHDs compared to DS individuals without CHDs. METHODS: We used the Illumina EPIC array and whole-genome bisulfite sequencing (WGBS) to quantitate DNA methylation for 86 NDBS samples from the California Biobank Program: (1) 45 DS-CHD (27 female, 18 male) and (2) 41 DS non-CHD (27 female, 14 male). We analyzed global CpG methylation and identified differentially methylated regions (DMRs) in DS-CHD versus DS non-CHD comparisons (both sex-combined and sex-stratified) corrected for sex, age of blood collection, and cell-type proportions. CHD DMRs were analyzed for enrichment in CpG and genic contexts, chromatin states, and histone modifications by genomic coordinates and for gene ontology enrichment by gene mapping. DMRs were also tested in a replication dataset and compared to methylation levels in DS versus typical development (TD) WGBS NDBS samples. RESULTS: We found global CpG hypomethylation in DS-CHD males compared to DS non-CHD males, which was attributable to elevated levels of nucleated red blood cells and not seen in females. At a regional level, we identified 58, 341, and 3938 CHD-associated DMRs in the Sex Combined, Females Only, and Males Only groups, respectively, and used machine learning algorithms to select 19 Males Only loci that could distinguish CHD from non-CHD. DMRs in all comparisons were enriched for gene exons, CpG islands, and bivalent chromatin and mapped to genes enriched for terms related to cardiac and immune functions. Lastly, a greater percent

BACKGROUND: Although recent sequencing studies have revealed that 10% of childhood gliomas are caused by rare germline mutations, the role of common variants is undetermined and no genome-wide significant risk loci for pediatric central nervous system tumors have been identified to date. METHODS: Meta-analysis of 3 population-based genome-wide association studies comprising 4069 children with glioma and 8778 controls of multiple genetic ancestries. Replication was performed in a separate case-control cohort. Quantitative trait loci analyses and a transcriptome-wide association study were conducted to assess possible links with brain tissue expression across 18 628 genes. RESULTS: Common variants in CDKN2B-AS1 at 9p21.3 were significantly associated with astrocytoma, the most common subtype of glioma in children (rs573687, P-value of 6.974e-10, OR 1.273, 95% CI 1.179-1.374). The association was driven by low-grade astrocytoma (P-value of 3.815e-9) and exhibited unidirectional effects across all 6 genetic ancestries. For glioma overall, the association approached genome-wide significance (rs3731239, P-value of 5.411e-8), while no significant association was observed for high-grade tumors. Predicted decreased brain tissue expression of CDKN2B was significantly associated with astrocytoma (P-value of 8.090e-8). CONCLUSIONS: In this population-based genome-wide association study meta-analysis, we identify and replicate 9p21.3 (CDKN2B-AS1) as a risk locus for childhood astrocytoma, thereby establishing the first genome-wide significant evidence of common variant predisposition in pediatric neuro-oncology. We furthermore provide a functional basis for the association by showing a possible link to decreased brain tissue CDKN2B expression and substantiate that genetic susceptibility differs between low- and high-grade astrocytoma.

BACKGROUND: Although recent sequencing studies have revealed that 10% of childhood gliomas are caused by rare germline mutations, the role of common variants is undetermined and no genome-wide significant risk loci for pediatric central nervous system tumors have been identified to date. METHODS: Meta-analysis of 3 population-based genome-wide association studies comprising 4069 children with glioma and 8778 controls of multiple genetic ancestries. Replication was performed in a separate case-control cohort. Quantitative trait loci analyses and a transcriptome-wide association study were conducted to assess possible links with brain tissue expression across 18 628 genes. RESULTS: Common variants in CDKN2B-AS1 at 9p21.3 were significantly associated with astrocytoma, the most common subtype of glioma in children (rs573687, P-value of 6.974e-10, OR 1.273, 95% CI 1.179-1.374). The association was driven by low-grade astrocytoma (P-value of 3.815e-9) and exhibited unidirectional effects across all 6 genetic ancestries. For glioma overall, the association approached genome-wide significance (rs3731239, P-value of 5.411e-8), while no significant association was observed for high-grade tumors. Predicted decreased brain tissue expression of CDKN2B was significantly associated with astrocytoma (P-value of 8.090e-8). CONCLUSIONS: In this population-based genome-wide association study meta-analysis, we identify and replicate 9p21.3 (CDKN2B-AS1) as a risk locus for childhood astrocytoma, thereby establishing the first genome-wide significant evidence of common variant predisposition in pediatric neuro-oncology. We furthermore provide a functional basis for the association by showing a possible link to decreased brain tissue CDKN2B expression and substantiate that genetic susceptibility differs between low- and high-grade astrocytoma.

Graph Convolutional Networks (GCNs) have been proved successful in the field of semi-supervised node classification by extracting structural information from graph data. However, the random selection of labeled nodes used by GCNs may lead to unstable generalization performance of GCNs. In this paper, we propose an efficient method for the deterministic selection of labeled nodes: the Determinate Node Selection (DNS) algorithm. The DNS algorithm identifies two categories of representative nodes in the graph: typical nodes and divergent nodes. These labeled nodes are selected by exploring the structure of the graph and determining the ability of the nodes to represent the distribution of data within the graph. The DNS algorithm can be applied quite simply on a wide range of semi-supervised graph neural network models for node classification tasks. Through extensive experimentation, we have demonstrated that the incorporation of the DNS algorithm leads to a remarkable improvement in the average accuracy of the model and a significant decrease in the standard deviation, as compared to the original method.

Recently, 3D object detection has attracted significant attention and achieved continuous improvement in real road scenarios. The environmental information is collected from a single sensor or multi-sensor fusion to detect interested objects. However, most of the current 3D object detection approaches focus on developing advanced network architectures to improve the detection precision of the object rather than considering the dynamic driving scenes, where data collected from sensors equipped in the vehicle contain various perturbation features. As a result, existing work cannot still tackle the perturbation issue. In order to solve this problem, we propose a group equivariant bird's eye view network (GeqBevNet) based on the group equivariant theory, which introduces the concept of group equivariant into the BEV fusion object detection network. The group equivariant network is embedded into the fused BEV feature map to facilitate the BEV-level rotational equivariant feature extraction, thus leading to lower average orientation error. In order to demonstrate the effectiveness of the GeqBevNet, the network is verified on the nuScenes validation dataset in which mAOE can be decreased to 0.325. Experimental results demonstrate that GeqBevNet can extract more rotational equivariant features in the 3D object detection of the actual road scene and improve the performance of object orientation prediction., Comment: 8 pages,3 figures

Long-tailed (LT) classification is an unavoidable and challenging problem in the real world. Most existing long-tailed classification methods focus only on solving the class-wise imbalance while ignoring the attribute-wise imbalance. The deviation of a classification model is caused by both class-wise and attribute-wise imbalance. Due to the fact that attributes are implicit in most datasets and the combination of attributes is complex, attribute-wise imbalance is more difficult to handle. For this purpose, we proposed a novel long-tailed classification framework, aiming to build a multi-granularity classification model by means of invariant feature learning. This method first unsupervisedly constructs Coarse-Grained forest (CLF) to better characterize the distribution of attributes within a class. Depending on the distribution of attributes, one can customize suitable sampling strategies to construct different imbalanced datasets. We then introduce multi-center loss (MCL) that aims to gradually eliminate confusing attributes during feature learning process. The proposed framework does not necessarily couple to a specific LT classification model structure and can be integrated with any existing LT method as an independent component. Extensive experiments show that our approach achieves state-of-the-art performance on both existing benchmarks ImageNet-GLT and MSCOCO-GLT and can improve the performance of existing LT methods. Our codes are available on GitHub: \url{https://github.com/jinyery/cognisance}, Comment: This is another research work to apply leading tree structure along with deep learning architecture, aiming to deal with attribute-wise long-tail distribution within class

We demonstrate an effective strategy to prohibit lithium dendrite formation in lithium metal batteries (LMBs) by forming a Li3N-rich protection solid electrolyte interphase (SEI) layer on a three-dimensional (3D) austenitic stainless steel (SS) surface, which maintains the structural stability during Li plating/stripping. The porous SS with 20% Cr is treated by ultra-low temperature ion-nitriding (250 ∼ 350 °C) for energy-saving, resulting in a lithiophilic nitride layer with a chromium nitride (CrN) phase on the surface, evidenced by the X-ray photoelectron spectroscopy (XPS) and further confirmed by Density functional theory (DFT) calculation. Upon cycling, it spontaneously reacts with metallic Li to form a strong and highly ionic conductive Li3N protection layer, which functionally inhibit continuous side reactions with electrolyte and facilitate the Li-ions transport, consequently leading to a flat Li deposition and ultralong cycling. Additionally, the 3D skeleton of SS reduces the areal current density and provides enough space to relieve volume expansion. The electrochemical results showed that the N-350-SS/Li (porous SS nitrided at 350℃ and composited with Li metal) anode in symmetrical cells exhibited extremely low voltage hysteresis of 14 mV and prolonged excellent cycling for over 2500 h in the normal LiPF6 electrolyte at 1 mA cm−2. The half cells retained an excellent Coulombic efficiency (CE) of above 98% and the full cells retained 98.0% of its initial capacity without much degradation after 200 cycles at 3C. This enhancement method is simple and low-cost and the SS can be successfully applied in the real high-energy–density LMBs.

Accelerated aging is a hallmark of Down syndrome (DS), with adults experiencing early-onset Alzheimer's disease and premature aging of the skin, hair, and immune and endocrine systems. Accelerated epigenetic aging has been found in the blood and brain tissue of adults with DS but when premature aging in DS begins remains unknown. We investigated whether accelerated aging in DS is already detectable in blood at birth. We assessed the association between age acceleration and DS using five epigenetic clocks in 346 newborns with DS and 567 newborns without DS using Illumina MethylationEPIC DNA methylation array data. We compared two epigenetic aging clocks (DNAmSkinBloodClock and pan-tissue DNAmAge) and three epigenetic gestational age clocks (Haftorn, Knight, and Bohlin) between DS and non-DS newborns using linear regression adjusting for observed age, sex, batch, deconvoluted blood cell proportions, and genetic ancestry. Targeted sequencing of GATA1 was performed in a subset of 184 newborns with DS to identify somatic mutations associated with transient abnormal myelopoiesis. DS was significantly associated with increased DNAmSkinBloodClock (effect estimate = 0.2442, p < 0.0001), with an epigenetic age acceleration of 244 days in newborns with DS after adjusting for potential confounding factors (95% confidence interval: 196-292 days). We also found evidence of epigenetic age acceleration associated with somatic GATA1 mutations among newborns with DS (p = 0.015). DS was not associated with epigenetic gestational age acceleration. We demonstrate that accelerated epigenetic aging in the blood of DS patients begins prenatally, with implications for the pathophysiology of immunosenescence and other aging-related traits in DS.

Genome-wide association studies have identified a growing number of single nucleotide polymorphisms (SNPs) associated with childhood acute lymphoblastic leukemia (ALL), yet the functional roles of most SNPs are unclear. Multiple lines of evidence suggest epigenetic mechanisms may mediate the impact of heritable genetic variation on phenotypes. Here, we investigated whether DNA methylation mediates the effect of genetic risk loci for childhood ALL. We performed an epigenome-wide association study (EWAS) including 808 childhood ALL cases and 919 controls from California-based studies using neonatal blood DNA. For differentially methylated CpG positions (DMPs), we next conducted association analysis with 23 known ALL risk SNPs followed by causal mediation analyses addressing the significant SNP-DMP pairs. DNA methylation at CpG cg01139861, in the promoter region of IKZF1, mediated the effects of the intronic IKZF1 risk SNP rs78396808, with the average causal mediation effect (ACME) explaining ~ 30% of the total effect (ACME P = 0.0031). In analyses stratified by self-reported race/ethnicity, the mediation effect was only significant in Latinos, explaining ~ 41% of the total effect of rs78396808 on ALL risk (ACME P = 0.0037). Conditional analyses confirmed the presence of at least three independent genetic risk loci for childhood ALL at IKZF1, with rs78396808 unique to non-European populations. We also demonstrated that the most significant DMP in the EWAS, CpG cg13344587 at gene ARID5B (P = 8.61x10-10), was entirely confounded by the ARID5B ALL risk SNP rs7090445. Our findings provide new insights into the functional pathways of ALL risk SNPs and the DNA methylation differences associated with risk of childhood ALL.

Accelerated aging is a hallmark of Down syndrome (DS), with adults experiencing early-onset Alzheimer's disease and premature aging of the skin, hair, and immune and endocrine systems. Accelerated epigenetic aging has been found in the blood and brain tissue of adults with DS but when premature aging in DS begins remains unknown. We investigated whether accelerated aging in DS is already detectable in blood at birth. We assessed the association between age acceleration and DS using five epigenetic clocks in 346 newborns with DS and 567 newborns without DS using Illumina MethylationEPIC DNA methylation array data. We compared two epigenetic aging clocks (DNAmSkinBloodClock and pan-tissue DNAmAge) and three epigenetic gestational age clocks (Haftorn, Knight, and Bohlin) between DS and non-DS newborns using linear regression adjusting for observed age, sex, batch, deconvoluted blood cell proportions, and genetic ancestry. Targeted sequencing of GATA1 was performed in a subset of 184 newborns with DS to identify somatic mutations associated with transient abnormal myelopoiesis. DS was significantly associated with increased DNAmSkinBloodClock (effect estimate = 0.2442, p < 0.0001), with an epigenetic age acceleration of 244 days in newborns with DS after adjusting for potential confounding factors (95% confidence interval: 196-292 days). We also found evidence of epigenetic age acceleration associated with somatic GATA1 mutations among newborns with DS (p = 0.015). DS was not associated with epigenetic gestational age acceleration. We demonstrate that accelerated epigenetic aging in the blood of DS patients begins prenatally, with implications for the pathophysiology of immunosenescence and other aging-related traits in DS.

Genome-wide association studies have identified a growing number of single nucleotide polymorphisms (SNPs) associated with childhood acute lymphoblastic leukemia (ALL), yet the functional roles of most SNPs are unclear. Multiple lines of evidence suggest that epigenetic mechanisms may mediate the impact of heritable genetic variation on phenotypes. Here, we investigated whether DNA methylation mediates the effect of genetic risk loci for childhood ALL. We performed an epigenome-wide association study (EWAS) including 808 childhood ALL cases and 919 controls from California-based studies using neonatal blood DNA. For differentially methylated CpG positions (DMPs), we next conducted association analysis with 23 known ALL risk SNPs followed by causal mediation analyses addressing the significant SNP-DMP pairs. DNA methylation at CpG cg01139861, in the promoter region of IKZF1, mediated the effects of the intronic IKZF1 risk SNP rs78396808, with the average causal mediation effect (ACME) explaining ~30% of the total effect (ACME P = 0.0031). In analyses stratified by self-reported race/ethnicity, the mediation effect was only significant in Latinos, explaining ~41% of the total effect of rs78396808 on ALL risk (ACME P = 0.0037). Conditional analyses confirmed the presence of at least three independent genetic risk loci for childhood ALL at IKZF1, with rs78396808 unique to non-European populations. We also demonstrated that the most significant DMP in the EWAS, CpG cg13344587 at gene ARID5B (P = 8.61 × 10-10), was entirely confounded by the ARID5B ALL risk SNP rs7090445. Our findings provide new insights into the functional pathways of ALL risk SNPs and the DNA methylation differences associated with risk of childhood ALL.

BACKGROUND: Periconceptional folate intake is associated with the establishment of DNA methylation in offspring; however, variations in this relation by food sources compared with folic acid supplements are not described. Also, maternal folate intake is associated with decreased risk of pediatric acute lymphoblastic leukemia (ALL), but the mechanism is not known. OBJECTIVES: We evaluated the relation between periconceptional folate intake by source and DNA methylation at birth in a cohort of pediatric ALL cases and controls in an epigenome-wide association study. METHODS: Genome-wide DNA methylation status obtained from archived neonatal blood spots from pediatric ALL cases (n = 189) and controls (n = 205) in the California Childhood Leukemia Study (CCLS) from 1995-2008 was compared with periconceptional folate from total, food, and supplemental sources using multivariable linear regression. Further stratification was performed by income, education, ethnicity, and total folate intake. We evaluated variable DNA methylation response to periconceptional folate by ALL case status through an interaction term. RESULTS: Two significant differentially methylated probes (DMPs) were associated with food and supplemental periconceptional folate intake in all subjects (n = 394). The top differentially methylated region at the promoter region of DUSP22(dual specificity phosphatase 22) demonstrated DNA hypermethylation in ALL cases but not in controls in response to total and food folate intake. We further identified 8 interaction term DMPs with variable DNA methylation response to folate intake by ALL case status. Further stratification of the cohort by education and ethnicity revealed a substantially higher number of DMPs associated with supplemental folic acid intake in Hispanic subjects with lower income and educational level. CONCLUSIONS: We identified modest associations between periconceptional folate intake and DNA methylation differing by so

In modern solid-state drives (SSDs), the indexing of flash pages is a critical component in their storage controllers. It not only affects the data access performance, but also determines the efficiency of the precious in-device DRAM resource. A variety of address mapping schemes and optimization techniques have been proposed. However, most of them were developed with human-driven heuristics. They cannot automatically capture diverse data access patterns at runtime in SSD controllers, which leaves a large room for improvement. In this paper, we present a learning-based flash translation layer (FTL), named LeaFTL, which learns the address mapping to tolerate dynamic data access patterns via linear regression at runtime. By grouping a large set of mapping entries into a learned segment, it significantly reduces the memory footprint of the address mapping table, which further benefits the data caching in SSD controllers. LeaFTL also employs various optimization techniques, including out-of-band metadata verification to tolerate mispredictions, optimized flash allocation, and dynamic compaction of learned index segments. We implement LeaFTL with an SSD simulator and evaluate it with various storage workloads. LeaFTL saves the memory consumption of the mapping table by 2.9x on average and improves the storage performance by 1.4x on average, in comparison with state-of-the-art FTL schemes., Comment: This paper is accepted at the 28th Conference on Architectural Support for Programming Languages and Operating Systems (ASPLOS 2023)

Previous works show that Pre-trained Language Models (PLMs) can capture factual knowledge. However, some analyses reveal that PLMs fail to perform it robustly, e.g., being sensitive to the changes of prompts when extracting factual knowledge. To mitigate this issue, we propose to let PLMs learn the deterministic relationship between the remaining context and the masked content. The deterministic relationship ensures that the masked factual content can be deterministically inferable based on the existing clues in the context. That would provide more stable patterns for PLMs to capture factual knowledge than randomly masking. Two pre-training tasks are further introduced to motivate PLMs to rely on the deterministic relationship when filling masks. Specifically, we use an external Knowledge Base (KB) to identify deterministic relationships and continuously pre-train PLMs with the proposed methods. The factual knowledge probing experiments indicate that the continuously pre-trained PLMs achieve better robustness in factual knowledge capturing. Further experiments on question-answering datasets show that trying to learn a deterministic relationship with the proposed methods can also help other knowledge-intensive tasks., Comment: Accepted at EMNLP 2022

The centrality and diversity of the labeled data are very influential to the performance of semi-supervised learning (SSL), but most SSL models select the labeled data randomly. This study first construct a leading forest that forms a partially ordered topological space in an unsupervised way, and select a group of most representative samples to label with one shot (differs from active learning essentially) using property of homeomorphism. Then a kernelized large margin metric is efficiently learned for the selected data to classify the remaining unlabeled sample. Optimal leading forest (OLF) has been observed to have the advantage of revealing the difference evolution along a path within a subtree. Therefore, we formulate an optimization problem based on OLF to select the samples. Also with OLF, the multiple local metrics learning is facilitated to address multi-modal and mix-modal problem in SSL, especially when the number of class is large. Attribute to this novel design, stableness and accuracy of the performance is significantly improved when compared with the state-of-the-art graph SSL methods. The extensive experimental studies have shown that the proposed method achieved encouraging accuracy and efficiency. Code has been made available at https://github.com/alanxuji/DeLaLA., Comment: 12 pages, ready to submit to a journal

This paper presents a systematic study on the security of modern file systems, following a vulnerability-centric perspective. Specifically, we collected 377 file system vulnerabilities committed to the CVE database in the past 20 years. We characterize them from four dimensions that include why the vulnerabilities appear, how the vulnerabilities can be exploited, what consequences can arise, and how the vulnerabilities are fixed. This way, we build a deep understanding of the attack surfaces faced by file systems, the threats imposed by the attack surfaces, and the good and bad practices in mitigating the attacks in file systems. We envision that our study will bring insights towards the future development of file systems, the enhancement of file system security, and the relevant vulnerability mitigating solutions., Comment: 20 pages, 8 figures

Recently, there has been a trend to investigate the factual knowledge captured by Pre-trained Language Models (PLMs). Many works show the PLMs' ability to fill in the missing factual words in cloze-style prompts such as "Dante was born in [MASK]." However, it is still a mystery how PLMs generate the results correctly: relying on effective clues or shortcut patterns? We try to answer this question by a causal-inspired analysis that quantitatively measures and evaluates the word-level patterns that PLMs depend on to generate the missing words. We check the words that have three typical associations with the missing words: knowledge-dependent, positionally close, and highly co-occurred. Our analysis shows: (1) PLMs generate the missing factual words more by the positionally close and highly co-occurred words than the knowledge-dependent words; (2) the dependence on the knowledge-dependent words is more effective than the positionally close and highly co-occurred words. Accordingly, we conclude that the PLMs capture the factual knowledge ineffectively because of depending on the inadequate associations., Comment: Accepted at Findings of ACL 2022

Vanadium diselenide (VSe2) exhibits versatile electronic and magnetic properties in the trigonal prismatic (H-) and octahedral (T-) phases. Compared to the metallic T-phase, the H-phase with a tunable semiconductor property is predicted to be a ferrovalley material with spontaneous valley polarization. Herein we report an epitaxial growth of the monolayer 2D VSe2 on a mica substrate via the chemical vapor deposition (CVD) method by introducing salt in the precursor. Our first-principles calculations suggest that the monolayer H-phase VSe2 with a large lateral size is thermodynamically favorable. The honeycomb-like structure and the broken symmetry are directly observed by spherical aberration-corrected scanning transmission electron microscopy (STEM) and confirmed by giant second harmonic generation (SHG) intensity. The p-type transport behavior is further evidenced by the temperature-dependent resistance and field-effect device study. The present work introduces a new phase-stable 2D transition metal dichalcogenide, opening the prospect of novel electronic and spintronics device design.

We demonstrate an effective strategy to prohibit lithium dendrite formation in lithium metal batteries (LMBs) by forming a Li3N-rich protection solid electrolyte interphase (SEI) layer on a three-dimensional (3D) austenitic stainless steel (SS) surface, which maintains the structural stability during Li plating/stripping. The porous SS with 20% Cr is treated by ultra-low temperature ion-nitriding (250 ∼ 350 °C) for energy-saving, resulting in a lithiophilic nitride layer with a chromium nitride (CrN) phase on the surface, evidenced by the X-ray photoelectron spectroscopy (XPS) and further confirmed by Density functional theory (DFT) calculation. Upon cycling, it spontaneously reacts with metallic Li to form a strong and highly ionic conductive Li3N protection layer, which functionally inhibit continuous side reactions with electrolyte and facilitate the Li-ions transport, consequently leading to a flat Li deposition and ultralong cycling. Additionally, the 3D skeleton of SS reduces the areal current density and provides enough space to relieve volume expansion. The electrochemical results showed that the N-350-SS/Li (porous SS nitrided at 350℃ and composited with Li metal) anode in symmetrical cells exhibited extremely low voltage hysteresis of 14 mV and prolonged excellent cycling for over 2500 h in the normal LiPF6 electrolyte at 1 mA cm−2. The half cells retained an excellent Coulombic efficiency (CE) of above 98% and the full cells retained 98.0% of its initial capacity without much degradation after 200 cycles at 3C. This enhancement method is simple and low-cost and the SS can be successfully applied in the real high-energy–density LMBs.

Down syndrome is associated with genome-wide perturbation of gene expression, which may be mediated by epigenetic changes. We perform an epigenome-wide association study on neonatal bloodspots comparing 196 newborns with Down syndrome and 439 newborns without Down syndrome, adjusting for cell-type heterogeneity, which identifies 652 epigenome-wide significant CpGs (P < 7.67 × 10-8) and 1,052 differentially methylated regions. Differential methylation at promoter/enhancer regions correlates with gene expression changes in Down syndrome versus non-Down syndrome fetal liver hematopoietic stem/progenitor cells (P < 0.0001). The top two differentially methylated regions overlap RUNX1 and FLI1, both important regulators of megakaryopoiesis and hematopoietic development, with significant hypermethylation at promoter regions of these two genes. Excluding Down syndrome newborns harboring preleukemic GATA1 mutations (N = 30), identified by targeted sequencing, has minimal impact on the epigenome-wide association study results. Down syndrome has profound, genome-wide effects on DNA methylation in hematopoietic cells in early life, which may contribute to the high frequency of hematological problems, including leukemia, in children with Down syndrome.

Medulloblastoma (MB) is a malignant pediatric brain tumor arising in the cerebellum. Although abnormal GABAergic receptor activation has been described in MB, studies have not yet elucidated the contribution of receptor-independent GABA metabolism to MB pathogenesis. We find primary MB tumors globally display decreased expression of GABA transaminase (ABAT), the protein responsible for GABA metabolism, compared with normal cerebellum. However, less aggressive WNT and SHH subtypes express higher ABAT levels compared with metastatic G3 and G4 tumors. We show that elevated ABAT expression results in increased GABA catabolism, decreased tumor cell proliferation, and induction of metabolic and histone characteristics mirroring GABAergic neurons. Our studies suggest ABAT expression fluctuates depending on metabolite changes in the tumor microenvironment, with nutrient-poor conditions upregulating ABAT expression. We find metastatic MB cells require ABAT to maintain viability in the metabolite-scarce cerebrospinal fluid by using GABA as an energy source substitute, thereby facilitating leptomeningeal metastasis formation.

Down syndrome is associated with genome-wide perturbation of gene expression, which may be mediated by epigenetic changes. We perform an epigenome-wide association study on neonatal bloodspots comparing 196 newborns with Down syndrome and 439 newborns without Down syndrome, adjusting for cell-type heterogeneity, which identifies 652 epigenome-wide significant CpGs (P < 7.67 × 10-8) and 1,052 differentially methylated regions. Differential methylation at promoter/enhancer regions correlates with gene expression changes in Down syndrome versus non-Down syndrome fetal liver hematopoietic stem/progenitor cells (P < 0.0001). The top two differentially methylated regions overlap RUNX1 and FLI1, both important regulators of megakaryopoiesis and hematopoietic development, with significant hypermethylation at promoter regions of these two genes. Excluding Down syndrome newborns harboring preleukemic GATA1 mutations (N = 30), identified by targeted sequencing, has minimal impact on the epigenome-wide association study results. Down syndrome has profound, genome-wide effects on DNA methylation in hematopoietic cells in early life, which may contribute to the high frequency of hematological problems, including leukemia, in children with Down syndrome.

Medulloblastoma (MB) is a malignant pediatric brain tumor arising in the cerebellum. Although abnormal GABAergic receptor activation has been described in MB, studies have not yet elucidated the contribution of receptor-independent GABA metabolism to MB pathogenesis. We find primary MB tumors globally display decreased expression of GABA transaminase (ABAT), the protein responsible for GABA metabolism, compared with normal cerebellum. However, less aggressive WNT and SHH subtypes express higher ABAT levels compared with metastatic G3 and G4 tumors. We show that elevated ABAT expression results in increased GABA catabolism, decreased tumor cell proliferation, and induction of metabolic and histone characteristics mirroring GABAergic neurons. Our studies suggest ABAT expression fluctuates depending on metabolite changes in the tumor microenvironment, with nutrient-poor conditions upregulating ABAT expression. We find metastatic MB cells require ABAT to maintain viability in the metabolite-scarce cerebrospinal fluid by using GABA as an energy source substitute, thereby facilitating leptomeningeal metastasis formation.

Machine reading comprehension (MRC) is a challenging natural language processing (NLP) task. Recently, the emergence of pre-trained models (PTM) has brought this research field into a new era, in which the training objective plays a key role. The masked language model (MLM) is a self-supervised training objective that widely used in various PTMs. With the development of training objectives, many variants of MLM have been proposed, such as whole word masking, entity masking, phrase masking, span masking, and so on. In different MLM, the length of the masked tokens is different. Similarly, in different machine reading comprehension tasks, the length of the answer is also different, and the answer is often a word, phrase, or sentence. Thus, in MRC tasks with different answer lengths, whether the length of MLM is related to performance is a question worth studying. If this hypothesis is true, it can guide us how to pre-train the MLM model with a relatively suitable mask length distribution for MRC task. In this paper, we try to uncover how much of MLM's success in the machine reading comprehension tasks comes from the correlation between masking length distribution and answer length in MRC dataset. In order to address this issue, herein, (1) we propose four MRC tasks with different answer length distributions, namely short span extraction task, long span extraction task, short multiple-choice cloze task, long multiple-choice cloze task; (2) four Chinese MRC datasets are created for these tasks; (3) we also have pre-trained four masked language models according to the answer length distributions of these datasets; (4) ablation experiments are conducted on the datasets to verify our hypothesis. The experimental results demonstrate that our hypothesis is true.

With the development of information technology, human beings are constantly producing a large amount of information at all times. How to obtain the information that users are interested in from the large amount of information has become an issue of great concern to users and even business managers. In order to solve this problem, from traditional machine learning to deep learning recommendation systems, researchers continue to improve optimization models and explore solutions. Because researchers have optimized more on the recommendation model network structure, they have less research on enriching recommendation model features, and there is still room for in-depth recommendation model optimization. Based on the DIN\cite{Authors01} model, this paper adds multi-head and multi-modal modules, which enriches the feature sets that the model can use, and at the same time strengthens the cross-combination and fitting capabilities of the model. Experiments show that the multi-head multi-modal DIN improves the recommendation prediction effect, and outperforms current state-of-the-art methods on various comprehensive indicators.

With the improvement of living standards, user requirements of modern products are becoming increasingly more diversified and personalized. Traditional product design methods can no longer satisfy the market needs due to their strong subjectivity, small survey scope, poor real-time data, and lack of visual display, which calls for the development of big data driven product design methodology. Big data in the product lifecycle contains valuable information for guiding product design, such as customer preferences, market demands, product evaluation, and visual display: online product reviews reflect customer evaluations and requirements; product images contain information of shape,color, and texture which can inspire designers to get initial design schemes more quickly or even directly generate new product images. How to efficiently collect product design related data and exploit them effectively during the whole product design process is thus critical to modern product design. This paper aims to conduct a comprehensive survey on big data driven product design. It will help researchers and practitioners to comprehend the latest development of relevant studies and applications centered on how big data can be processed, analyzed, and exploited in aiding product design. We first introduce several representative traditional product design methods and highlight their limitations. Then we discuss current and potential applications of textual data, image data, audio data, and video data in product design cycles. Finally, major deficiencies of existing data driven product design studies and future research directions are summarized. We believe that this study can draw increasing attention to modern data driven product design., Comment: 33 pages

Human-designed rules are widely used to build industry applications. However, it is infeasible to maintain thousands of such hand-crafted rules. So it is very important to integrate the rule knowledge into neural networks to build a hybrid model that achieves better performance. Specifically, the human-designed rules are formulated as Regular Expressions (REs), from which the equivalent Minimal Deterministic Finite Automatons (MDFAs) are constructed. We propose to use the MDFA as an intermediate model to capture the matched RE patterns as rule-based features for each input sentence and introduce these additional features into neural networks. We evaluate the proposed method on the ATIS intent classification task. The experiment results show that the proposed method achieves the best performance compared to neural networks and four other methods that combine REs and neural networks when the training dataset is relatively small.

Machine Reading Comprehension (MRC) is a challenging Natural Language Processing(NLP) research field with wide real-world applications. The great progress of this field in recent years is mainly due to the emergence of large-scale datasets and deep learning. At present, a lot of MRC models have already surpassed human performance on various benchmark datasets despite the obvious giant gap between existing MRC models and genuine human-level reading comprehension. This shows the need for improving existing datasets, evaluation metrics, and models to move current MRC models toward "real" understanding. To address the current lack of comprehensive survey of existing MRC tasks, evaluation metrics, and datasets, herein, (1) we analyze 57 MRC tasks and datasets and propose a more precise classification method of MRC tasks with 4 different attributes; (2) we summarized 9 evaluation metrics of MRC tasks, 7 attributes and 10 characteristics of MRC datasets; (3) We also discuss key open issues in MRC research and highlighted future research directions. In addition, we have collected, organized, and published our data on the companion website(https://mrc-datasets.github.io/) where MRC researchers could directly access each MRC dataset, papers, baseline projects, and the leaderboard., Comment: 68 pages

Collecting supporting evidence from large corpora of text (e.g., Wikipedia) is of great challenge for open-domain Question Answering (QA). Especially, for multi-hop open-domain QA, scattered evidence pieces are required to be gathered together to support the answer extraction. In this paper, we propose a new retrieval target, hop, to collect the hidden reasoning evidence from Wikipedia for complex question answering. Specifically, the hop in this paper is defined as the combination of a hyperlink and the corresponding outbound link document. The hyperlink is encoded as the mention embedding which models the structured knowledge of how the outbound link entity is mentioned in the textual context, and the corresponding outbound link document is encoded as the document embedding representing the unstructured knowledge within it. Accordingly, we build HopRetriever which retrieves hops over Wikipedia to answer complex questions. Experiments on the HotpotQA dataset demonstrate that HopRetriever outperforms previously published evidence retrieval methods by large margins. Moreover, our approach also yields quantifiable interpretations of the evidence collection process., Comment: Accepted at AAAI 2021

A major challenge in materials design is how to efficiently search the vast chemical design space to find the materials with desired properties. One effective strategy is to develop sampling algorithms that can exploit both explicit chemical knowledge and implicit composition rules embodied in the large materials database. Here, we propose a generative machine learning model (MatGAN) based on a generative adversarial network (GAN) for efficient generation of new hypothetical inorganic materials. Trained with materials from the ICSD database, our GAN model can generate hypothetical materials not existing in the training dataset, reaching a novelty of 92.53% when generating 2 million samples. The percentage of chemically valid (charge neutral and electronegativity balanced) samples out of all generated ones reaches 84.5% by our GAN when trained with materials from ICSD even though no such chemical rules are explicitly enforced in our GAN model, indicating its capability to learn implicit chemical composition rules. Our algorithm could be used to speed up inverse design or computational screening of inorganic materials., Comment: 15 pages