EPCO-29. EPIGENOMICS OF THE GLIOMA LONGITUDINAL ANALYSIS (GLASS) CONSORTIUM

Adult diffuse gliomas are central nervous system (CNS) tumors that arise from the malignant transformation of glial cells. Nearly all gliomas will recur despite standard treatment however, current histopathological grading fails to predict which of them will relapse and/or progress. The Glioma Longitudinal AnalySiS (GLASS) consortium is a large-scale collaboration that aims to investigate the molecular profiling of matched primary and recurrent glioma samples from multiple institutions in order to better understand the dynamic evolution of these tumors. At this time, the cohort comprises 946 samples across 11 institutions and among those, 864 have DNA methylation data available. The current molecular classification based on 7 subtypes published by TCGA in 2016 was applied to the dataset. Among the IDH wildtype tumors, 33% (16/49) of the patients showed a change of subtype upon recurrence, whereas most of them (9/16) were Classic-like at the primary stage but changed to either Mesenchymal-like or PA-like at the recurrent level. Among the IDH mutant tumors, 15% (22/142) showed a change of subtype at recurrent stage, in which 16 out of 22 progressed from G-CIMP-high to G-CIMP-low. Although some tumors progressed to a different subtype upon recurrence, an unsupervised analysis showed that the samples tend to cluster by patient instead of by subtype. By estimating the copy number alterations of these tumors using DNA methylation, the overall copy number profile of the recurrent samples remains similar to their primary counterpart. From this initial analysis using epigenomic data, we were able to characterize some aspects of glioma evolution and how the DNA methylation is associated with the progression of these tumors to different subtypes. These findings corroborate the importance of epigenetics in gliomas and can potentially lead to the identification of new biomarkers that can reflect tumor burden and predict its development.

Medical Care of Adults With Down Syndrome: A Clinical Guideline

IMPORTANCE

Down syndrome is the most common chromosomal condition, and average life expectancy has increased substantially, from 25 years in 1983 to 60 years in 2020. Despite the unique clinical comorbidities among adults with Down syndrome, there are no clinical guidelines for the care of these patients.

OBJECTIVE

To develop an evidence-based clinical practice guideline for adults with Down syndrome.

EVIDENCE REVIEW

The Global Down Syndrome Foundation Medical Care Guidelines for Adults with Down Syndrome Workgroup (n = 13) developed 10 Population/Intervention/ Comparison/Outcome (PICO) questions for adults with Down syndrome addressing multiple clinical areas including mental health (2 questions), dementia, screening or treatment of diabetes, cardiovascular disease, obesity, osteoporosis, atlantoaxial instability, thyroid disease, and celiac disease. These questions guided the literature search in MEDLINE, EMBASE, PubMed, PsychINFO, Cochrane Library, and the TRIP Database, searched from January 1, 2000, to February 26, 2018, with an updated search through August 6, 2020. Using the GRADE (Grading of Recommendations, Assessment, Development, and Evaluation) methodology and the Evidence-to-Decision framework, in January 2019, the 13-member Workgroup and 16 additional clinical and scientific experts, nurses, patient representatives, and a methodologist developed clinical recommendations. A statement of good practice was made when there was a high level of certainty that the recommendation would do more good than harm, but there was little direct evidence.

FINDINGS

From 11 295 literature citations associated with 10 PICO questions, 20 relevant studies were identified. An updated search identified 2 additional studies, for a total of 22 included studies (3 systematic reviews, 19 primary studies), which were reviewed and synthesized. Based on this analysis, 14 recommendations and 4 statements of good practice were developed. Overall, the evidence base was limited. Only 1 strong recommendation was formulated: screening for Alzheimer-type dementia starting at age 40 years. Four recommendations (managing risk factors for cardiovascular disease and stroke prevention, screening for obesity, and evaluation for secondary causes of osteoporosis) agreed with existing guidance for individuals without Down syndrome. Two recommendations for diabetes screening recommend earlier initiation of screening and at shorter intervals given the high prevalence and earlier onset in adults with Down syndrome.

CONCLUSIONS AND RELEVANCE

These evidence-based clinical guidelines provide recommendations to support primary care of adults with Down syndrome. The lack of high-quality evidence limits the strength of the recommendations and highlights the need for additional research.

Microbial selenium metabolism: a brief history, biogeochemistry and ecophysiology

Selenium is an essential trace element for organisms from all three domains of life. Microorganisms, in particular, mediate reductive transformations of selenium that govern the element's mobility and bioavailability in terrestrial and aquatic environments. Selenium metabolism is not just ubiquitous but an ancient feature of life likely extending back to the universal common ancestor of all cellular lineages. As with the sulfur biogeochemical cycle, reductive transformations of selenium serve two metabolic functions: assimilation into macromolecules and dissimilatory reduction during anaerobic respiration. This review begins with a historical overview of how research in both aspects of selenium metabolism has developed. We then provide an overview of the global selenium biogeochemical cycle, emphasizing the central role of microorganisms in the cycle. This serves as a basis for a robust discussion of current models for the evolution of the selenium biogeochemical cycle over geologic time, and how knowledge of the evolution and ecophysiology of selenium metabolism can enrich and refine these models. We conclude with a discussion of the ecophysiological function of selenium-respiring prokaryotes within the cycle, and the tantalizing possibility of oxidative selenium transformations during chemolithoautotrophic growth.

Ultrasound Shear Wave Elastography to Assess Tissue Mechanical Properties in Somatic Dysfunction: A Feasibility Study

CONTEXT

Diagnosis of somatic dysfunction is based on subjective palpatory osteopathic assessments. This subjectivity has posed a challenge for researchers in studying osteopathic medicine. The development and use of radiological imaging techniques to objectively confirm or quantify muscle tissue stiffness associated with somatic dysfunction could be of benefit in osteopathic clinical practice, training, and further research.

OBJECTIVES

To investigate the feasibility of ultrasound shear wave elastography (SWE) to quantify muscle tissue stiffness in somatic dysfunction before and after osteopathic manipulative treatment (OMT).

METHODS

In this prospective study, we assessed lumbar spine somatic dysfunction in 20 adult patients before and after a single OMT session using standard osteopathic palpatory assessments by osteopathic physician faculty members in the Department of Osteopathic Principle and Practice at Rocky Vista University College of Osteopathic Medicine (Utah campus). Shear wave velocity (SWV, m/s) was measured in lumbar paraspinal muscle tissue using a commercial ultrasonography scanner on all participants immediately before and after OMT. In this study, OMT techniques targeted the iliocostalis lumborum and included the articulatory technique, balanced ligamentous tension, facilitated positional release, high-velocity, low-amplitude technique, muscle energy, myofascial release, and the Still technique at the discretion of the osteopathic physician. The difference in SWV between muscle tissues with and without dysfunction, and differences in SWV of dysfunctional tissue before and after OMT were examined using unpaired and paired t tests, as appropriate. The correlation between SWV measurements and osteopathic assessments was examined by the Spearman rank correlation. Intra- and interobserver reliability was analyzed using intraclass correlation coefficient.

RESULTS

The difference in SWV between muscle tissues with and without somatic dysfunction was significant before OMT (mean [SD], 1.93 [0.44] vs 1.69 [0.19]; P=.03) and was not significant after OMT (mean [SD], 1.69 [0.19] vs 1.53 [0.31]; P=.05). The difference in SWV in the same tissue with somatic dysfunction before and after OMT was significant (mean [SD], 1.93 [0.44] vs 1.52 [0.3]; P<.001). The SWV value highly correlated with manual osteopathic assessments (r=0.72). Intra- and interobserver reliability for performing SWE in somatic dysfunction was good (intraclass correlation coefficient >0.80).

CONCLUSIONS

The results of this study show that ultrasound SWE can objectively assess muscle tissue stiffness for diagnosis of somatic dysfunctions and for muscle tissue stiffness changes after OMT.

Abstract A10: Glioma cell-free DNA methylation marker for diagnosis and monitoring

Genome-wide DNA-methylation profiling has shown that epigenetic abnormalities are biologically and clinically important signs in glioma and can be used to classify these tumors into distinct prognostic groups. Thus far, DNA methylation profiling of gliomas has required surgically resected tissue. Because gliomas release tumoral material into biofluids, such as blood, we developed a method using a liquid biopsy (LB) for minimally invasive surveillance of the tumor. We performed a genome-wide CpG methylation profile of cell-free DNA extracted from serum of patients with primary glioma (N=48) and during clinical follow-up (N=33). We used a supervised machine learning (ML) approach to identify glioma-specific epigenetic signatures in the LB (eLB) (N=1000, false-discovery rate &lt; 0.003) to distinguish patients with glioma from those without glioma. The epigenetic profiles of matched tumor tissue showed that there was 67% concordance between the eLB signatures of the glioma in serum and tissue. Using these signatures as input for ML, we derived the glioma epigenetic liquid biopsy (GeLB) score. An independent validation study (68 gliomas and 65 nongliomas) showed that a GeLB score ≥ 50% could predict with 96% accuracy (sensitivity = 97.1% and specificity = 95.9%) whether a patient had glioma or not. Changes in GeLB score also reflected clinicopathologic changes observed during surveillance (e.g., progression, pseudoprogression, or response to a clinical trial drug). Our results suggest that the GeLB score could be used as a complementary approach to diagnose glioma and follow up patients postsurgically. Finally, we provide an online tool for predicting glioma based on the cell-free DNA methylome (hbtc.shinyapps.io/GeLB/).

Citation Format: Thais Sabedot, Tathiane Malta, James Snyder, Kevin Nelson, Michael Wells, Ana deCarvalho, Abir Mukherjee, Dhan Chitale, Maritza Mosella, Karam Asmaro, Adam Robin, Mark Rosenblum, Tom Mikkelsen, Jack Rock, Laila Poisson, Ian Lee, Tobias Walbert, Steven Kalkanis, Ana Valeria Castro, Houtan Noushmehr. Glioma cell-free DNA methylation marker for diagnosis and monitoring [abstract]. In: Proceedings of the AACR Special Conference on Advances in Liquid Biopsies; Jan 13-16, 2020; Miami, FL. Philadelphia (PA): AACR; Clin Cancer Res 2020;26(11_Suppl):Abstract nr A10.

Abstract A11: Serum cell-free DNA methylome-based signatures distinguish pituitary tumor from other neoplasias and by clinicopathologic features

Background: Several reports have indicated that distinct epigenomic patterns of pituitary tumor (PT) tissue, specifically DNA methylation, distinguish these tumors according to their functionality and could be involved in their pathogenesis. Thus far, molecular diagnosis and classification systems that guide clinical management of these tumors rely on the tissue profiling obtained by invasive surgical approaches (e.g., excision). However, increasing evidence showed that CNS tumors release cell material into the circulation, despite the existence of the blood-brain barrier, creating an opportunity for molecular profiling using a liquid biopsy. Although the pituitary portal system and the invasion of the cavernous system by PT may facilitate the spillage of tumor cell material into the bloodstream, until now liquid biopsy has not been described in pituitary tumors. Considering the stability, the cell specificity, and the reported role of DNA methylation in PT, we investigated the feasibility to detect and define PT-specific methylation-based signatures in the serum of patients harboring these tumors.

Methods and Findings: In order to identify PT-specific methylation-based signatures in the serum of patients with PT, we conducted unsupervised and supervised analysis of the methylome profile of paired serum cfDNA (EPIC array) and/or tissue from 13 patients with pituitary macroadenomas (9 males; median age: 62; 9 nonfunctioning/4 functioning, 6 invasive/7 noninvasive), 4 controls serum (nontumor and healthy), and patients with other CNS tumors or conditions (114 gliomas, 6 meningiomas, 1 brain metastasis, 1 colloid cyst, 6 radiation necrosis). Serum PT-tissue-specific methylation was also used as input into a machine learning algorithm to generate a PT score, which was tested in an independent cohort. Unsupervised and supervised analysis indicated that the serum methylome from patients harboring PT was distinct from controls and other CNS diseases. A generated PT score using the 37 identified PT tissue-specific methylation probes was able to distinguish serum from patients harboring PT from other CNS tumors. Serum-derived PT function-specific DMP also distinguished PT according to functional status. Serum-derived invasive-specific DMPs were less prominent to define invasive status in an independent cohort of PT, most probably due to the limited number of cases.

Conclusion: This is the first study to show the feasibility of profiling methylome in the serum of patients with PT using cfDNA. In addition, we identified unique methylation signatures that distinguished PT from other CNS tumors and according to distinct PT subtypes. These results underpin the potential role of methylation profile and liquid biopsy as a noninvasive approach to assess clinically relevant molecular features that can be used as diagnostic, prognostic, and surveillance in patients with PT.

Citation Format: Michael Wells, Karam Asmaro, Jack Rock, Thais Sabedot, Maritza Mosella, Tathiane Malta, Kevin Nelson, James Snyder, Steven Kalkanis, Ana Valeria Castro, Houtan Noushmehr. Serum cell-free DNA methylome-based signatures distinguish pituitary tumor from other neoplasias and by clinicopathologic features [abstract]. In: Proceedings of the AACR Special Conference on Advances in Liquid Biopsies; Jan 13-16, 2020; Miami, FL. Philadelphia (PA): AACR; Clin Cancer Res 2020;26(11_Suppl):Abstract nr A11.

A National Surgical Quality Improvement Program Analysis of Postoperative Major and Minor Complications in Patients with Spinal Metastatic Disease

BACKGROUND

Major complications after spine metastasis surgery are prioritized in the literature with little consideration of the more frequent minor events such as pneumonia or urinary tract infection. We analyzed incidence and risk factors of postsurgical complications in patients with spinal metastasis extracted from the National Surgical Quality Improvement Program (NSQIP). We also developed a useful predictive model to estimate the probability of occurrence of complications.

METHODS

A total of 1176 patients diagnosed with spinal metastasis were extracted from NSQIP. Variables screened included age, sex, tumor location, patient's functional status, comorbidities, laboratory values, and case urgency. Two multivariate logistic regression models were designed to evaluate risk factors and likelihood of event occurrence.

RESULTS

Minor events occurred twice as frequently compared with major complications (36% vs. 18% of patients). The most common major event was death (10%); the most frequent minor event was need for postoperative transfusion (29.4%). In the multivariate analysis, elderly age, emergency case, preoperative leukocytosis, and smoking status retained significance for major complications; American Society of Anesthesiologists classes 4-5, low hematocrit levels, and intradural extramedullary location of the tumor retained significance for minor complications. The predictive models designed explained 72% of the variability in major complications occurrence and 67% for minor events.

CONCLUSIONS

Smoking status and emergent surgery were found to be the strongest independent predictors of major complications, whereas higher American Society of Anesthesiologists class showed a greater association with minor events. The predictive models produced can be a useful aid for surgeons to identify those patients who are at greater risk of developing postoperative adverse events.

OR32-03 Serum Cell-Free Methylation-Based Signatures Distinguishes Pituitary Tumors According to Functional Status and from Other Neoplasia: A Liquid Biopsy Approach

BACKGROUND: Several reports have indicated that distinct epigenomic patterns of pituitary tumors (PT), specifically DNA methylation, distinguish these tumor tissues according to their functionality and could be involved in their pathogenesis. Thus far, molecular diagnosis and classification criteria that guide clinical management of these tumors rely on the tissue profiling obtained by invasive surgical approaches (e.g. excision). However, increasing evidence confirmed that central nervous system (CNS) tumors release cell material into the circulation creating an opportunity for molecular profiling of these tumors using a blood-based liquid biopsy. Considering that 1) the pituitary portal system and the invasion of the cavernous system by PT may facilitate the spillage of tumor cell material into the bloodstream and 2) the stability, cell-specificity and reportedly the role of DNA methylation in PT, we hypothesized that liquid biopsy would be feasible to detect and define specific methylation-based signatures in the serum of patients harboring PT. Methods and Findings: We conducted analyses of the methylomes of paired serum circulating cell-free DNA (cfDNA) and tumor tissue from patients harboring PT (EPIC array) to identify serum-derived pituitary tumor-specific methylation-based signatures (sPTMet n=37) in a cohort comprised by 13 patients with pituitary macroadenomas (9 males; median age: 62; 9 Nonfunctioning/4functioning, 6 invasive/7noninvasive), 4 controls (non-tumor) and patients with other CNS tumors or conditions (114 gliomas, 6 meningiomas, 1 brain metastasis, 1 colloid cyst, 6 radiation necrosis). Unsupervised and supervised analysis indicated that the serum methylome from patients harboring PT was distinct from controls and other CNS diseases. Using the sPTMet as input into a machine learning algorithm, we generated a PT score that classified the serum of an independent cohort as PT or non-PT, with high accuracy. We identified serum-derived differentially methylated probes (DMP, n=3288) that distinguished PT according to their function (functioning and nonfunctioning). When overlapped with an independent cohort, these DMP also distinguished PT tissue according to their functional status. Conclusion: Our results showed the feasibility to identify PT-specific methylation signatures by profiling the methylome of serum cfDNA from patients with PT. These signatures distinguished PT from other CNS tumors and according to their subtypes. These results underpin the potential role of methylation profile and liquid biopsy as a noninvasive approach to assess clinically relevant molecular features. Potentially, tumor-specific serum-derived methylation signature may be used as a diagnostic, prognostic and surveillance tool as well to identify actionable molecular markers in patients with PT.

28 Can Comprehensive Geriatric Assessment be Achieved in the Emergency Department?

Background

Our National Health Service is facing unprecedented challenges to accommodate our frailer healthcare users. The gold standard tool for the identification and management of frailty is the Comprehensive Geriatric Assessment (CGA) and has been shown to lead to better outcomes in terms of morbidity and mortality.

Introduction

With a largely elderly demographic profile in the East of England, the Norfolk and Norwich University Hospital opened the first Older People’s Emergency Department (OPED) in the UK in 2017. This work reviews the effectiveness of a geriatrician-led CGA in a dedicated OPED, which operates during daylight hours, compared to usual care in Accident & Emergency (A&E).

Methods

99 patients assessed in OPED and 99 patients assessed overnight in A&E during February 2019 were included in this retrospective study. Electronic case notes for each patient were reviewed by the authors and results were expressed as percentages.

Results

OPED outperformed A&E in all components of the CGA; strongest areas included assessing for pain, falls risk and activities of daily living. Both departments performed well in reviewing medications and assessing for safeguarding concerns. Areas for improvement include assessing for mood disorders, sensory impairment, discussing Do Not Attempt Cardiopulmonary Resuscitation status, and end of life care plans. The average length of stay of OPED patients was only 7.3 days compared to 8.7 days in A&E, and 89% of OPED patients were discharged back to their usual residences compared to 87% in A&E.

Conclusions

The improved CGA process in OPED has led to better outcomes, notably through a reduction in the average length of inpatient stay. Nevertheless, certain components of the CGA still require improvement. Further examination is needed to assess long-term mortality to support the use of CGA in the emergency setting.

Loss of the Epigenetic Mark 5-hmC in Psoriasis: Implications for Epidermal Stem Cell Dysregulation

Epigenetic regulation has a profound influence on stem cell fate during normal development in maintenance of physiologic tissue homeostasis. Here we report diminished ten-eleven translocation (TET) methylcytosine dioxygenase expression and loss of the DNA hydroxymethylation mark 5-hydroxymethylcytosine (5-hmC) in keratinocyte stem cells and transit amplifying cells in human psoriasis and in imiquimod-induced murine psoriasis. Loss of 5-hmC was associated with dysregulated keratinocyte stem cell kinetics, resulting in accumulation of nestin and FABP5-expressing transit amplifying cells to produce classic psoriatic epidermal architecture. Moreover, 5-hmC loss was accompanied by diminished TET1 and TET2 mRNA expression. Genome-wide mapping of epidermal 5-hmC in murine psoriasis revealed loci-specific loss of 5-hmC in genes regulating stem cell homeostasis, including MBD1, RTN1, STRN4, PRKD2, AKT1, and MAPKAP2, as well as those associated with RAR and Wnt/β-catenin signaling pathways. In vitro restoration of TET expression by ascorbic acid was accomplished in cultured human keratinocyte stem cells to show similar Ca⁺⁺-induced differentiation, resulting in increased 5-hmC levels and reduced nestin expression. To our knowledge, an epigenetic deficiency in psoriasis with relevance to stem cell dysregulation has not been previously reported. This observation raises the possibility that epigenetic modifiers that impact on the TET-5-hmC pathway may be a relevant approach of heretofore unappreciated therapeutic utility.

Identification of novel dynamin-related protein 1 (Drp1) GTPase inhibitors: Therapeutic potential of Drpitor1 and Drpitor1a in cancer and cardiac ischemia-reperfusion injury

Mitochondrial fission is important in physiological processes, including coordination of mitochondrial and nuclear division during mitosis, and pathologic processes, such as the production of reactive oxygen species (ROS) during cardiac ischemia-reperfusion injury (IR). Mitochondrial fission is mainly mediated by dynamin-related protein 1 (Drp1), a large GTPase. The GTPase activity of Drp1 is essential for its fissogenic activity. Therefore, we aimed to identify Drp1 inhibitors and evaluate their anti-neoplastic and cardioprotective properties in five cancer cell lines (A549, SK-MES-1, SK-LU-1, SW 900, and MCF7) and an experimental cardiac IR injury model. Virtual screening of a chemical library revealed 17 compounds with high predicted affinity to the GTPase domain of Drp1. In silico screening identified an ellipticine compound, Drpitor1, as a putative, potent Drp1 inhibitor. We also synthesized a congener of Drpitor1 to remove the methoxymethyl group and reduce hydrolytic lability (Drpitor1a). Drpitor1 and Drpitor1a inhibited the GTPase activity of Drp1 without inhibiting the GTPase of dynamin 1. Drpitor1 and Drpitor1a have greater potency than the current standard Drp1 GTPase inhibitor, mdivi-1, (IC50 for mitochondrial fragmentation are 0.09, 0.06, and 10 μM, respectively). Both Drpitors reduced proliferation and induced apoptosis in cancer cells. Drpitor1a suppressed lung cancer tumor growth in a mouse xenograft model. Drpitor1a also inhibited mitochondrial ROS production, prevented mitochondrial fission, and improved right ventricular diastolic dysfunction during IR injury. In conclusion, Drpitors are useful tools for understanding mitochondrial dynamics and have therapeutic potential in treating cancer and cardiac IR injury.

Determination of Domoic Acid Toxins in Shellfish by Biosense ASP ELISAA Direct Competitive Enzyme-Linked Immunosorbent Assay: Collaborative Study

A collaborative study was conducted on the Biosense amnesic shellfish poisoning (ASP) enzyme-linked immunosorbent assay (ELISA) for the determination of domoic acid (DA) toxins in shellfish in order to obtain interlaboratory validation data for the method. In addition, a method comparison study was performed to evaluate the ASP ELISA as an alternative to the current liquid chromatography (LC) reference method for DA determination. The study material comprised 16 shellfish samples, including blue mussels, Pacific oysters, and king scallops, spiked with contaminated mussel homogenates to contain 0.120 mg DA/kg shellfish flesh. The shellfish samples were extracted with 50% aqueous methanol, and the supernatants were directly analyzed. Sixteen participating laboratories in 10 countries reported data from the ASP ELISA, and 4 of these laboratories also reported data from instrumental LC analysis. The participating laboratories achieved interlaboratory precision estimates for the 8 Youden paired shellfish samples in the range of 1020% for RSDr (mean 14.8 4%), and 1329% for RSDR (mean 22.7 6%). The precision estimates for the ELISA data did not show a strong dependence on the DA concentration in the study samples, and the overall precision achieved was within the acceptable range of the Horwitz guideline with HorRat values ranging from 1.1 to 2.4 (mean HorRat 1.7 0.5). The analysis of shellfish samples spiked with certified reference material (CRM)-ASP-MUS-b gave recoveries in the range of 88122%, with an average recovery of 104 10%. The estimate on method accuracy was supported by a correlation slope of 1.015 (R2 = 0.992) for the determined versus the expected DA values. Furthermore, the correlation of the ASP ELISA results with those for the instrumental LC analyses of the same sample extracts gave a correlation slope of 1.29 (R2 = 0.984). This indicates some overestimation of DA levels in shellfish by the ELISA, but it is also a result of apparent low recoveries for the LC methods. This interlaboratory study demonstrates that the ASP ELISA is suitable for the routine determination and monitoring of DA toxins in shellfish, and that it offers a rapid and cost-effective methodology with high sample throughput.

INNV-15. CLINICAL DATA THAT MATTERS: A DISTILLATION OF NEURO-ONCOLOGY CLINICAL TRIAL INCLUSION CRITERIA USING MACHINE LEARNING

INTRODUCTION

Neuro-oncologic conditions have dismal outcomes, ineffective treatments, poor access to clinical trials, and variability in care. Clinical trials do not capture a patient’s complete journey and are restricted to select populations. ‘Real-world-evidence’ (RWE) attempts to inform point of care decisions through routine collection of data with a clinical-trial-like rigor. RWE complements existing knowledge through broad patient participation, collection throughout disease course, and creation of large multidimensional datasets “knowledge network of disease” 1,2. RWE implementation is hindered by unstructured data, uncertainty of relevant features, and semantic heterogeneity. Clinical attributes were selected from trial inclusion criteria and prioritized for structuring in clinic notes for abstraction.

METHOD

We queried Clinicaltrials.gov from 1/1/2018-12/31/2018, refined to North America, recruiting, interventional, and adult. Meningioma, pituitary, glioblastoma, astrocytoma, oligodendroglioma, and ependymoma were chosen based on incidence3. Lymphoma and nerve sheath tumors were omitted. “Brain tumor” and “glioma” were added. ‘K-nearest-neighbor’ tokenization parsed inclusion criteria4. Document term matrix (n-gram) converted text to vectors5. A generative probabilistic model using ‘Latent Dirichlet Allocation’ plotted words into 10 clusters6. Hierarchal clustering was used to compare histology with terms.

RESULTS

401 trials parsed into 3676 statements and 4008 keywords. 10 clusters of terms were similarly distributed amongst histologies, suggesting generalizability across tumor types. Cluster revealed 8 categories: 1) Time: enrollment; 2) Performance status: KPS; 3) Testing: mutations, upper limit of normal, routine hematologic laboratory assays; 4) Imaging: extent of surgery; 5) Pregnancy/childbearing; 6) Tumor grade; 7) Treatment history: recurrence, chemotherapy, radiation, time; 8) Informed consent

CONCLUSIONS

Dissecting the compendium of clinical trials using machine learning can identify general parameters for trial enrollment to guide RWE clinical collection. Using practical definitions of the most germane trial data, specific information can be sought after and defined to improve research quality, maximize research yields and improve patient care whilst minimizing wasted research and clinical endeavors.

GENE-24. DNA METHYLATION SIGNATURES DETECTED IN A SERUM-BASED LIQUID BIOPSY DISTINGUISH FUNCTIONAL AND INVASIVENESS FEATURES IN PITUITARY ADENOMAS

Pituitary tumors are the second most common CNS neoplasia (~15%). Despite mostly benign and slow-growing, they may be nonfunctioning and invade surrounding structures resulting in significant comorbidities. Currently, classification of PT according to their risk for aggressiveness is mainly based on invasiveness detected by imaging methods and histopathological features which requires surgically resected tumor. Being able to detect molecular markers associated with tumor subtypes and behavior pre-surgically using minimally invasive approaches (blood draw) is desirable in these tumors and may help to address current diagnostic and therapeutic challenges. In tissue specimens, distinct DNA methylation patterns distinguish PT according to their functional status but their role in invasiveness is still unclear. We hypothesized that profiling cell-free DNA (cfDNA) released by PT into the bloodstream allow the identification of epigenetic markers associated with relevant clinicopathological features. Genome-wide methylome profile of paired serum cfDNA (EPIC array) and tissue from 13 patients with pituitary macroadenomas (9 males; median age: 62; 9 Nonfunctioning/4 functioning, 6 invasive/7 noninvasive) and 3 controls serum (patients with epilepsy). Unsupervised analysis of the serum methylome from patients harboring PT was distinct from controls and other diseases (hypopituitarism, glioma and colorectal cancer) and supervised analysis (Wilcoxon Rank-sum Test) identified significant differentially methylated probes (DMP) that segregated PT from control serum specimens. Nonfunctioning and invasive-specific DMPs identified in the serum also defined functional, and less prominently invasive status, in the tissue of an independent cohort of PT. This is the first study to show the feasibility to profile the serum methylome from patients with PT using cfDNA. In addition, we identified unique methylation signatures that distinguished PT according to functional and invasiveness subtypes. These results underpin the potential role of methylation profile and liquid biopsy as a noninvasive approach to assess clinically relevant molecular features in the serum of patients harboring PT.

P01.02 Serum-derived DNA methylation markers distinguish functional and invasiveness subtypes in patients harboring pituitary tumors

BACKGROUND

Molecular profiling of circulating biomarkers released by tumors has a relevant clinical value in central nervous system (CNS) tumors, but its feasibility has not been investigated in pituitary tumors (PT) despite being the second common intraaxial tumors of the CNS (~15%). Although usually benign and slow-growing, they can be nonfunctioning and invade surrounding structures resulting in significant comorbidities. DNA methylation aberrations distinguish PT according to their functional status but their role in invasiveness is still unclear. Pre-surgical detection of clinically relevant molecular markers associated with tumor behavior can address current diagnostic and therapeutic challenges. We hypothesized that PT release cell-free DNA (cfDNA) into the bloodstream allowing for the profiling of epigenetic markers associated with relevant clinicopathological features.

MATERIAL AND METHODS

Genome-wide methylome profile of paired serum cfDNA (EPIC array) and tissue from 13 patients with pituitary macroadenomas (9 males; median age: 62; 9 NFPT, 6 invasive) and 3 controls serum (patients with epilepsy).

RESULTS

Unsupervised analysis of the serum methylome from patients harboring PT was distinct from controls and other diseases (hypopituitarism, glioma and colorectal cancer) and supervised analysis (Wilcoxon Rank-sum Test) identified significant differentially methylated probes (DMP) that segregated PT from control serum specimens. Nonfunctioning and invasive-specific DMPs identified in the serum also defined functional, and less prominently invasive status, in the tissue of an independent cohort of PT.

CONCLUSION

This is the first study to show the feasibility to profile the methylome in the serum of patients with PT using cfDNA. In addition, we identified unique methylation signatures that distinguished PT according to functional and invasiveness subtypes. These results underpin the potential role of methylation profile and liquid biopsy as a noninvasive approach to assess clinically relevant molecular features in the serum of patients harboring PT.

OS1.5 Detection of glioma and prognostic subtypes by non-invasive circulating cell-free DNA methylation markers

BACKGROUND: Genome-wide DNA methylation profiling has shown that epigenetic abnormalities are biologically important in glioma and can be used to classify these tumors into distinct prognostic groups. Thus far, DNA profiling has required surgically resected glioma tissue; however, gliomas release tumoral material into biofluids providing an opportunity for a minimally invasive testing. While prior studies have shown that molecular markers can be detected in liquid biopsy (LB), there has been low sensitivity for tumor-specific markers. We hypothesize that the low sensitivity is due to the targeted assay methods. METHODS: Genome-wide CpG methylation levels in DNA of tumor tissue and cell-free DNA serum of glioma patients. RESULTS: We defined glioma-specific and IDH-specific epigenetic LB (eLB) signatures (Glioma-eLB and IDH-eLB, respectively) from serum cell-free DNA from patients diagnosed with glioma (N=15 IDH mutant and N=7 IDH wildtype) and with epilepsy (N=3). The epigenetic profiles of the matched tissue demonstrate that these eLB signatures reflected the signature of the tumor. Through cross-validation we show that Glioma-eLB can accurately predict a patient’s glioma from those with other neoplasias (N=6 Colon; N=14 Pituitary; N=3 Breast; N=4 Lung), non-neoplastic immunological conditions (N=22 sepsis; N=9 pancreatic islet transplantation), and from healthy individuals (sensitivity: 98%; specificity: 99%). Finally, IDH-eLB includes promoter methylated markers associated with genes known to be involved in glioma tumorigenesis (PVT1 and CXCR6). CONCLUSIONS: The application of the non-invasive eLB signature discovered in this study has the potential to complement the standard of care for patients harboring glioma.

This project is supported by the Henry Ford Health System, Department of Neurosurgery and the Hermelin Brain Tumor Center Foundation (A30935), United States National Institutes of Health (R01CA222146), and United States Department of Defense (CA170278)

OS8.8 Modified chromatin histone marks at noncoding elements are associated with aggressive meningioma subtype

Background: Although most meningiomas are non-malignant, there is a high recurrence rate among atypical and anaplastic (malignant) meningiomas (grades II/III). In addition, malignant meningioma usually progresses after treatment. Recently, based on DNA methylation, two subgroups of meningioma were described with recurrence-free survival differences (favorable and unfavorable). Epigenetic deregulation at distinct genomic elements such as enhancers can drive changes in gene expression and alter the transcriptional profile of the cancer cells. We seek to understand the mechanisms of meningioma recurrence and progression after initial treatment. Material and Methods: Two favorable and two unfavorable meningioma high grade tumor tissue were selected for this pilot study. Immunoprecipitation (Antibodies) for two different histone modifications (H3K4me3 and H3K27ac) were used to generate genome-wide chromatin-IP sequencing. Genome-wide DNA methylation profile was assessed for 90 meningioma cases including the four samples used in this pilot study.

Results: Using a FDR cutoff of 5%, we identified 10,049 H3K27ac and 5,752 H3K4me3 chromatin marks that distinguish favorable from unfavorable meningioma. We dichotomized the results into genomic regions overlapping known gene promoters and non-promoters. Promoter associated chromatin changes (H3K27ac and H3K4me3) coincide with DNA methylation changes in aggressive meningioma. The top 1,000 H3K27ac (sorted by fold difference) without H3K4me3 coincide with highly conserved DNA elements known to be associated with enhancers. These enhancers can drive expression of multiple genes simultaneously. Using DNA methylation differences between favorable and unfavorable meningioma, we will overlap with our unfavorable associated enhancers in order to refine our candidate enhancers for follow up mechanistic studies using aggressive meningioma cell lines.

Conclusion: Our preliminary results are the first to unravel the genome-wide chromatin changes associated with unfavorable or clinically aggressive meningioma. Identification of these candidate enhancers will provide knowledge of the role of epigenomics in the development of malignant meningioma and of opportunities for targeted therapy.

This project is supported by the Henry Ford Health System, Department of Neurosurgery and the Hermelin Brain Tumor Center Foundation (A30935), United States National Institutes of Health (R01CA222146), and United States Department of Defense (CA170278)

MLTI-05. IDENTIFYING BRAIN METASTATIC CASES FROM FREE TEXT CLINICAL NARRATIVES WITH REFINEMENT OF SEMANTIC HETEROGENEITY USING MACHINE LEARNING

INTRODUCTION: Brain metastatic disease (BM) is ripe for discovery using computational tools like machine learning (ML) due to disease complexity and multidimensional critical data (imaging, genomics, primary disease, drug exposures)¹. Leveraging real-world-evidence’ (RWE) from routine health data to inform clinical management is hindered by fragmented unstructured data and semantic heterogeneity². Clinical data in EHR and institutional registries are typically free text narratives absent common data elements (CDE). Curating existing data into CDE with machine learning (ML) may inform contemporary approaches (RWE, N-of-1 trials, and precision medicine) that are dependent on large high-quality datasets. Harvesting existing institutional registries may expand demographic representation, confirm benchmarks of established treatments, and provide test environment for prospective ML applications. METHOD: An R-based deep convoluted neural network (DNN) using keras and an API for Tensorflow python was trained on physician narratives of 2000 BM cases and 8000 other CNS conditions labeled by diagnosis spanning 17 years^3,4. The ML model was tested with 405 non-labeled narratives to: A) Identify BM from other CNS conditions (i.e. glioma, meningioma, non-tumor). B) Evaluate word embedding using GLoVe⁵ to standardize abbreviations and misspellings by assigning terms to CDE by training the model to plot “mets”, “metastases” and “spine” with the 20 most similar contextual words. RESULTS: DNN architecture achieved 97% accuracy in distinguishing BM (n=178) for others (n=227). “Mets” and “metastasis” have a connected contextual network suggesting shared meaning, whereas spine did not share a network. CONCLUSIONS: ML can identify BM cases in free-text registries which can serve as a quality control measure and aid data aggregation. Standardizing shorthand terminology to CDE with DNN trained in word embedding can possibly address semantic heterogeneity and facilitate data automation. Solutions are needed to compile and automate quality BM data across institutions to achieve the volume and complexity required for contemporary analysis using ML.