GMQN: A Reference-Based Method for Correcting Batch Effects and Probe Bias in HumanMethylation BeadChip

The Illumina HumanMethylation BeadChip is one of the most cost-effective methods to quantify DNA methylation levels at single-base resolution across the human genome, which makes it a routine platform for epigenome-wide association studies. It has accumulated tens of thousands of DNA methylation array samples in public databases, providing great support for data integration and further analysis. However, the majority of public DNA methylation data are deposited as processed data without background probes which are widely used in data normalization. Here, we present Gaussian mixture quantile normalization (GMQN), a reference based method for correcting batch effects as well as probe bias in the HumanMethylation BeadChip. Availability and implementation: https://github.com/MengweiLi-project/gmqn.

The ENmix DNA methylation analysis pipeline for Illumina BeadChip and comparisons with seven other preprocessing pipelines

Background Illumina DNA methylation arrays are high-throughput platforms for cost-effective genome-wide profiling of individual CpGs. Experimental and technical factors introduce appreciable measurement variation, some of which can be mitigated by careful “preprocessing” of raw data. Methods Here we describe the ENmix preprocessing pipeline and compare it to a set of seven published alternative pipelines (ChAMP, Illumina, SWAN, Funnorm, Noob, wateRmelon, and RnBeads). We use two large sets of duplicate sample measurements with 450 K and EPIC arrays, along with mixtures of isogenic methylated and unmethylated cell line DNA to compare raw data and that preprocessed via different pipelines. Results Our evaluations show that the ENmix pipeline performs the best with significantly higher correlation and lower absolute difference between duplicate pairs, higher intraclass correlation coefficients (ICC) and smaller deviations from expected methylation level in mixture experiments. In addition to the pipeline function, ENmix software provides an integrated set of functions for reading in raw data files from mouse and human arrays, quality control, data preprocessing, visualization, detection of differentially methylated regions (DMRs), estimation of cell type proportions, and calculation of methylation age clocks. ENmix is computationally efficient, flexible and allows parallel computing. To facilitate further evaluations, we make all datasets and evaluation code publicly available. Conclusion Careful selection of robust data preprocessing methods is critical for DNA methylation array studies. ENmix outperformed other pipelines in our evaluations to minimize experimental variation and to improve data quality and study power.

Clinically aggressive pediatric spinal ependymoma with novel MYC amplification demonstrates molecular and histopathologic similarity to newly described MYCN-amplified spinal ependymomas

Primary spinal cord tumors contribute to ≤ 10% of central nervous system tumors in individuals of pediatric or adolescent age. Among intramedullary tumors, spinal ependymomas make up ~ 30% of this rare tumor population. A twelve-year-old male presented with an intradural, extramedullary mass occupying the dorsal spinal canal from C6 through T2. Gross total resection and histopathology revealed a World Health Organization (WHO) grade 2 ependymoma. He recurred eleven months later with extension from C2 through T1-T2. Subtotal resection was achieved followed by focal proton beam irradiation and chemotherapy. Histopathology was consistent with WHO grade 3 ependymoma. Molecular profiling of the primary and recurrent tumors revealed a novel amplification of the MYC (8q24) gene, which was confirmed by fluorescence in situ hybridization studies. Although MYC amplification in spinal ependymoma is exceedingly rare, a newly described classification of spinal ependymoma harboring MYCN (2p24) amplification (SP-MYCN) has been defined by DNA methylation-array based profiling. These individuals typically present with a malignant progression and dismal outcomes, contrary to the universally excellent survival outcomes seen in other spinal ependymomas. DNA methylation array-based classification confidently classified this tumor as SP-MYCN ependymoma. Notably, among the cohort of 52 tumors comprising the SP-MYCN methylation class, none harbor MYC amplification, highlighting the rarity of this genomic amplification in spinal ependymoma. A literature review comparing our individual to reported SP-MYCN tumors (n = 26) revealed similarities in clinical, histopathologic, and molecular features. Thus, we provide evidence from a single case to support the inclusion of MYC amplified spinal ependymoma within the molecular subgroup of SP-MYCN.

Infrequent RAS mutation is not associated with specific histological phenotype in gliomas

Background Mutations in driver genes such as IDH and BRAF have been identified in gliomas. Meanwhile, dysregulations in the p53, RB1, and MAPK and/or PI3K pathways are involved in the molecular pathogenesis of glioblastoma. RAS family genes activate MAPK through activation of RAF and PI3K to promote cell proliferation. RAS mutations are a well-known driver of mutation in many types of cancers, but knowledge of their significance for glioma is insufficient. The purpose of this study was to reveal the frequency and the clinical phenotype of RAS mutant in gliomas. Methods This study analysed RAS mutations and their clinical significance in 242 gliomas that were stored as unfixed or cryopreserved specimens removed at Kyoto University and Osaka National Hospital between May 2006 and October 2017. The hot spots mutation of IDH1/2, H3F3A, HIST1H3B, and TERT promoter and exon 2 and exon 3 of KRAS, HRAS, and NRAS were analysed with Sanger sequencing method, and 1p/19q codeletion was analysed with multiplex ligation-dependent probe amplification. DNA methylation array was performed in some RAS mutant tumours to improve accuracy of diagnosis. Results RAS mutations were identified in four gliomas with three KRAS mutations and one NRAS mutation in one anaplastic oligodendroglioma, two anaplastic astrocytomas (IDH wild-type in each), and one ganglioglioma. RAS-mutant gliomas were identified with various types of glioma histology. Conclusion RAS mutation appears infrequent, and it is not associated with any specific histological phenotype of glioma.

GMQN: A reference-based method for correcting batch effects as well as probes bias in HumanMethylation BeadChip

Illumina HumanMethylation BeadChip is one of the most cost-effective ways to quantify DNA methylation levels at the single-base level across the human genome, which makes it a routine platform for epigenome-wide association studies. It has accumulated tens of thousands of DNA methylation array samples in public databases, thus provide great support for data integration and further analysis. However, majority of public DNA methylation data are deposited as processed data without background probes which are widely used in data normalization. Here we present Gaussian mixture quantile normalization (GMQN), a reference based method for correcting batch effects as well as probes bias in HumanMethylation BeadChip. Availability and implementation: https://github.com/MengweiLi-project/gmqn.

1455Leveraging multi-omic negative controls for effect estimation in molecular epidemiologic studies: A simulation study

Background Exploratory null-hypothesis significance testing (e.g. GWAS, EWAS) form the backbone of molecular epidemiology, however methods to identify true causal signals are underdeveloped. Via plasmode simulation, I evaluate two approaches to quantitatively control for shared unmeasured confounding and recover unbiased effects using complementary epigenomes and biologically-informed structural assumptions. Methods I adapt proposed negative control-based estimators, the control outcome calibration approach (COCA) and proximal g-computation (PG) to case studies in perinatal molecular epidemiology. COCA may be employed when maternal epigenome has no direct effects on phenotype and proxy shared unmeasured confounders and PG further with suitable genetic instruments (e.g. mQTLs). Baseline covariates were extracted from 777 mother-child pairs in a birth cohort with maternal blood and fetal cord DNA methylation array data. Treatment and outcome values were simulated in 2000 bootstraps. Bootstrapped, ordinary (COCA) and 2-stage (PG) least squares were fitted to estimate treatment effects and standard errors under various common settings of missing confounders (e.g. paternal data). Doubly-robust, machine learning estimators were explored. Results COCA and PG performed well in simplistic data generating processes. However, in real-world cohort simulations, COCA performed acceptably only in settings with strong proxy confounders, but otherwise poorly (median bias 610%; coverage 29%). PG performed slightly better. Alternatively, simple covariate adjustment for maternal methylation outperformed (median bias 22%; 71% coverage) COCA, PG, and machine learning estimators. Discussion Molecular epidemiology provides key opportunity to leverage biological knowledge against unmeasured confounding. Negative control calibration or adjustments may help under limited scenarios where assumptions are fulfilled, but should be tested with suitable simulations. Key messages Quantitative approaches for unmeasured confounding in molecular epidemiology are a critical gap. Negative control calibration or adjustment may help under limiting scenarios. Proposed estimators should be tested in simulation settings that closely mimic target data.

Examining the Vanishing Twin Hypothesis of Neural Tube Defects: Application of an Epigenetic Predictor for Monozygotic Twinning

Strong associations between neural tube defects (NTDs) and monozygotic (MZ) twinning have long been noted, and it has been suggested that NTD cases who do not present as MZ twins may be the survivors of MZ twinning events. We have recently shown that MZ twins carry a strong, distinctive DNA methylation signature and have developed an algorithm based on genomewide DNA methylation array data that distinguishes MZ twins from dizygotic twins and other relatives at well above chance level. We have applied this algorithm to published methylation data from five fetal tissues (placental chorionic villi, kidney, spinal cord, brain and muscle) collected from spina bifida cases (n = 22), anencephalic cases (n = 15) and controls (n = 19). We see no difference in signature between cases and controls, providing no support for a common etiological role of MZ twinning in NTDs. The strong associations therefore continue to await elucidation.

An analytical pipeline for DNA Methylation Array Biomarker Studies

DNA methylation is one of the most commonly studied epigenetic biomarkers, due to its role in disease and development. The Illumina Infinium methylation arrays still remains the most common method to interrogate methylation across the human genome, due to its capabilities of screening over 480, 000 loci simultaneously. As such, initiatives such as The Cancer Genome Atlas (TCGA) have utilized this technology to examine the methylation profile of over 20,000 cancer samples. There is a growing body of methods for pre-processing, normalisation and analysis of array-based DNA methylation data. However, the shape and sampling distribution of probe-wise methylation that could influence the way data should be examined was rarely discussed. Therefore, this article introduces a pipeline that predicts the shape and distribution of normalised methylation patterns prior to selection of the most optimal inferential statistics screen for differential methylation. Additionally, we put forward an alternative pipeline, which employed feature selection, and demonstrate its ability to select for biomarkers with outstanding differences in methylation, which does not require the predetermination of the shape or distribution of the data of interest. Availability: The Distribution test and the feature selection pipelines are available for download at: https://github.com/uqjlu8/DistributionTest Keywords: DNA methylation, Biomarkers, Cancers, Data Distribution, TCGA, 450K

EPCT-06. PRECISION ONCOLOGY IN THE PEDIATRIC TARGETED THERAPY 2.0 PROGRAM

Introduction Precise diagnoses and robust detection of actionable alterations is required for individualized treatments. By using extended molecular diagnostics, the Pediatric Targeted Therapy (PTT) 2.0 program aims at the improvement of diagnostic accuracy and detection of actionable alterations for pediatric high-risk patients. The impact of these analyses on clinical management is reported. Methods Pediatric patients with relapsed or progressive tumors after standard of care treatment were included, independent of histological diagnosis. Formalin fixed paraffin embedded material and a blood sample for germline correction were requested. DNA methylation array, targeted gene panel sequencing (130 genes), RNA and Sanger sequencing in selected cases, and immunohistochemistry (IHC) of selected markers (pERK, pAKT, pS6, PD-L1) were performed. A questionnaire-based follow-up was used to determine the clinical impact of the analysis. Results We enrolled n=263 patients from February 2017 to February 2019. Complete molecular analysis was possible for n=260 cases (99%). The most common entities were brain tumors (n=172/260, 65%). In brain tumors, DNA methylation array alone allowed robust diagnostic classification (score of >=0.9) in n=104/172 cases (60%). Actionable targets as detected by copy number calculation, gene panel sequencing, RNA sequencing and IHC were found in n=94/172 (55%) brain tumor cases. The most common actionable targets in brain tumors were MAPK (pERK, BRAF fusions, BRAF V600E), mTOR (pS6), PI3K (pAKT), CDK4/6 (CDKN2A/B loss), and immune checkpoints (PD-L1). Pathogenic germline alterations with clinical relevance were identified in n=12/172 brain tumor cases (6.9%) and were confirmed by Sanger sequencing, 5/12 (41%) of which were previously unknown. Clinical follow-up of subsequent treatment and outcome are ongoing. Conclusion The combination of next-generation diagnostics such as methylation arrays and targeted sequencing in addition to selected IHC markers added robust information with regard to diagnosis and actionable alterations. The impact on clinical decision-making and on outcome is currently being evaluated.