Large Public Data Research Articles

A cross-validation statistical framework for asymmetric data integration.

The proliferation of biobanks and large public clinical data sets enables their integration with a smaller amount of locally gathered data for the purposes of parameter estimation and model prediction. However, public data sets may be subject to context-dependent confounders and the protocols behind their generation are often opaque; naively integrating all external data sets equally can bias estimates and lead to spurious conclusions. Weighted data integration is a potential solution, but current methods still require subjective specifications of weights and can become computationally intractable. Under the assumption that local data are generated from the set of unknown true parameters, we propose a novel weighted integration method based upon using the external data to minimize the local data leave-one-out cross validation (LOOCV) error. We demonstrate how the optimization of LOOCV errors for linear and Cox proportional hazards models can be rewritten as functions of external data set integration weights. Significant reductions in estimation error and prediction error are shown using simulation studies mimicking the heterogeneity of clinical data as well as a real-world example using kidney transplant patients from the Scientific Registry of TransplantRecipients.

A Set of Comprehensive Evaluation System for Different Data Augmentation Methods

Data augmentation is an effective method to prevent model overfitting in deep learning, especially in medical image classification where data samples are small and difficult to obtain. In recent years, different data augmentation methods, such as those based on single data transformation, multiple data mixing, and learning data distribution, have been proposed one after another, but there has never been a systematic system to evaluate various data augmentation methods. An impartial and comprehensive data augmentation evaluation system not only can assess the benefits and drawbacks of existing augmentation approaches in a specific medical image classification but also can provide an effective research direction for the subsequent proposal of new medical image data augmentation methods, thereby advancing the development of auxiliary diagnosis technology based on medical images. Therefore, this paper proposes an objective and universal evaluation system for different data augmentation methods. In this method, different augmented methods are evaluated objectively and comprehensively in terms of classification accuracy and data diversity by using existing large public data sets. The method is universal and easy to operate. To imitate the prevalent small-sized data sets in deep learning, an equal-interval sampling technique based on similarity ranking is presented to select samples from large public data sets and construct a subset that can fully reflect the original set. The augmented data sets are then created using various data augmentation approaches based on the small-sized data sets. Finally, different data augmentation strategies are objectively and fully evaluated based on the comprehensive scores of classification accuracy and data diversity following data augmentation. The validity and feasibility of the suggested sampling method and assessment system in this study are demonstrated by experimental findings on numerous data sets.

Statistical kinematic modelling: concepts and model validity

Data reduction techniques are applied to reduce the volume of data while maintaining its integrity. For cyclic motion data, a reliable overview comparing these methods is lacking. Therefore, this study aims to evaluate the features of the different data reduction techniques by applying them to large public data sets. The periodicity of cyclic motion can be exploited by either analysing a single cycle or studying a series of cycles. Analysing single cycles requires a pre-processing step to isolate the amplitude variability. Three different alignment techniques were evaluated, namely Linear length normalisation (LLN), piecewise LLN (PLLN) and continuous registration (CR). CR showed to remove the most phase variation. For the data reduction, three techniques were assessed (i.e., principal component analysis (PCA), principal polynomial analysis (PPA) and multivariate functional PCA (MFPCA)) based on the in- and out-of-sample error, the compactness and the computation time. The differences were found to be minimal. From our results, PPA appeared to be most useful for data compression. Further, we recommend PCA and MFPCA for classification and feature extraction purposes. We suggest the use of PCA when computation time is key and we advise the use of MFPCA when the inclusion of different data sources is desired. In contrast, the analysis of a series of cycles requires a pre-processing step to decompose the series. Further, a regression model was used to compensate for the difference in fundamental frequency. PCA on FC and MFPCA with splines were applied on the frequency compensated curves. Both methods performed as good.

Research on origin‐destination travel demand prediction method of inter‐regional online taxi based on SpatialOD‐BiConvLSTM

In order to avoid empty or crowded trips of online taxis and shorten the distance between empty taxis and potential passengers, after adopting the Origin-Destination (OD) big data of taxis and considering geographic location information, cross-regional driving demand, as well as passenger origin demand and destination demand, a method combining SpatialOD and Bidirectional ConvLSTM (BiConvLSTM) neural network model is proposed to predict online taxi travel demand. The network model in the method takes into account the historical and future states of the data, and can extract the time and space characteristics of the data. A custom structural similarity loss function is used in the model training to measure the difference between the data. Through experiments on a large public taxi travel data set, the results show that the SMAPE error of this method is 23.83%, which is better than the existing OD demand forecasting methods and has strong practicability.

The Prognostic Value of Calcium in Post-Cardiovascular Surgery Patients in the Intensive Care Unit.

Background: Present researches exploring the prognostic value of calcium concentration are undermined by sample size and study design. Our study investigated the association of both total calcium (tCa) and ionized Ca (iCa) to short- and long-term mortality and other outcomes in post-cardiovascular surgery (PCS) patients admitted to intensive care unit (ICU) from two large public data sets.Methods: The Medical Information Mart for Intensive Care III (MIMIC-III) database and the eICU Collaborative Research Database (eICU) were inspected to identify PCS patients. The primary outcome was 28-day mortality. Multivariate regression was used to elucidate the relationship between calcium concentration and outcomes. The propensity score estimation was performed to validate our findings.Results: A total of 6122 and 914 patients were included from the MIMIC III and eICU data sets, respectively. The groups with the most patients were the mild hypo-iCa and hypo-tCa groups. The mild hypo-iCa group showed significant association with worse short-term and long-term prognosis, less use of ventilation, longer ICU and hospital stay, and more incidence of 7-day acute kidney injury.Conclusions: The mild hypo-iCa (0.9–1.15 mmol/L) within the first day of admission to the ICU could serve as an independent prognosis factor for PCS patients.

Simons Collaborative Marine Atlas Project (Simons CMAP): An open‐source portal to share, visualize, and analyze ocean data

AbstractSimons Collaborative Marine Atlas Project (Simons CMAP) is an open‐source data portal that interconnects large, complex, and diverse public data sets currently dispersed in different formats across different Oceanography discipline‐specific databases. Simons CMAP is designed to streamline the retrieval of custom subsets of data, the generation of data visualizations, and the analyses of diverse data, thus expanding the power of these potentially underutilized data sets for cross‐disciplinary studies of ocean processes. We describe a unified architecture that allows numerical model outputs, satellite products, and field observations to be readily shared, mined, and integrated regardless of data set size or resolution. A current focus of Simons CMAP is integration of physical, chemical, and biological data sets essential for characterizing the biogeography of key marine microbes across ocean basins and seasonal cycles. Using a practical example, we demonstrate how our unifying data harmonization plans significantly simplifies and allows for systematic data integration across all Simons CMAP data sets.

The Envirome Web Service: Patient context at the point of care

Patient context – the “envirome” – can have a significant impact on patient health. While envirome indicators are available through large scale public data sources, they are not provided in a format that can be easily accessed and interpreted at the point of care by healthcare providers with limited time during a patient encounter. We developed a clinical decision support tool to bring envirome indicators to the point of care in a large pediatric hospital system in the Kansas City region. The Envirome Web Service (EWS) securely geocodes patient addresses in real time to link their records with publicly available context data. End-users guided the design of the EWS, which presents summaries of patient context data in the electronic health record (EHR) without disrupting the provider workflow. Through surveys, focus groups, and a formal review by hospital staff, the EWS was deployed into production use, integrating publicly available data on food access with the hospital EHR. Evaluation of EWS usage during the 2020 calendar year shows that 1,034 providers viewed the EWS, with a total of 29,165 sessions. This suggests that the EWS was successfully integrated with the EHR and is highly visible. The results also indicate that 63 (6.1%) of the providers are regular users that opt to maintain the EWS in their custom workflows, logging more than 100 EWS sessions during the year. The vendor agnostic design of the EWS supports interoperability and makes it accessible to health systems with disparate EHR vendors.

ACTOR: a latent Dirichlet model to compare expressed isoform proportions to a reference panel.

The relative proportion of RNA isoforms expressed for a given gene has been associated with disease states in cancer, retinal diseases, and neurological disorders. Examination of relative isoform proportions can help determine biological mechanisms, but such analyses often require a per-gene investigation of splicing patterns. Leveraging large public data sets produced by genomic consortia as a reference, one can compare splicing patterns in a data set of interest with those of a reference panel in which samples are divided into distinct groups, such as tissue of origin, or disease status. We propose A latent Dirichlet model to Compare expressed isoform proportions TO a Reference panel (ACTOR), a latent Dirichlet model with Dirichlet Multinomial observations to compare expressed isoform proportions in a data set to an independent reference panel. We use a variational Bayes procedure to estimate posterior distributions for the group membership of one or more samples. Using the Genotype-Tissue Expression project as a reference data set, we evaluate ACTOR on simulated and real RNA-seq data sets to determine tissue-type classifications of genes. ACTOR is publicly available as an R package at https://github.com/mccabes292/actor.

Construction of a Virtual Opioid Bioprofile: A Data-Driven QSAR Modeling Study to Identify New Analgesic Opioids

Compared to traditional experimental approaches, computational modeling is a promising strategy to efficiently prioritize new candidates with low cost. In this study, we developed a novel data mining and computational modeling workflow proven to be applicable by screening new analgesic opioids. To this end, a large opioid data set was used as the probe to automatically obtain bioassay data from the PubChem portal. There were 114 PubChem bioassays selected to build quantitative structure-activity relationship (QSAR) models based on the testing results across the probe compounds. The compounds tested in each bioassay were used to develop 12 models using the combination of three machine learning approaches and four types of chemical descriptors. The model performance was evaluated by the coefficient of determination (R 2) obtained from 5-fold cross-validation. In total, 49 models developed for 14 bioassays were selected based on the criteria and were identified to be mainly associated with binding affinities to different opioid receptors. The models for these 14 bioassays were further used to fill data gaps in the probe opioids data set and to predict general drug compounds in the DrugBank data set. This study provides a universal modeling strategy that can take advantage of large public data sets for computer-aided drug design (CADD).

Alignment and parameterization of single cycle motion data

Abstract Motion capturing systems produce a large amount of information on the motion of individuals. A growing number of data reduction techniques have been developed to reduce the amount of data while keeping relevant information. An overview that compares and identifies the advantages and disadvantages of these methods on cyclic motion data is, however, lacking. Therefore, this study aims to assess the features of different data reduction techniques by applying them to a large public gait data set. Due to the periodicity of cyclic data, an individual cycle can be isolated and analyzed. The analysis of single cycles requires preprocessing steps to segment and align the individual cycles. The latter is needed to isolate the amplitude variability. Three alignment procedures with different complexity, namely Linear Length Normalization (LLN), Piecewise LNN (PLLN) and Continuous Registration (CR), are assessed based on the amount of resulting variation. Subsequently three data reduction techniques (i.e. Principal Component Analysis (PCA), Principal Polynomial Analysis (PPA) and Multivariate Functional PCA (MFPCA)) are applied to the aligned single gait cycles. The data reduction techniques are evaluated based on the in-sample error, the out-of-sample error, the compactness and the computation time to produce a model. The curves aligned with CR have the lowest remaining variation and thus the lowest amount of remaining phase variation. The differences between the different data reduction techniques appear to be minimal. PPA shows to be the most compact and is therefore recommended when compactness is crucial and out-of-sample performance is less essential. The use of MFPCA is advised when one wants to include data from different sources. PCA is suggested when computation time is key.

467. Spatial-Temporal Prediction Model of COVID-19 Mortality across 62 Counties in New York State

BackgroundCOVID-19 pandemic has resulted in considerable morbidity and mortality. New York State (NY) is the hotspot with most coronavirus cases, while there are spatial/temporal variations. Yet, few examined county-level factors of mortality in COVID-19 patients in NY. Based on the sociological framework in health, this study links large and representative public data to understand COVID-19 mortality in NY over different stages of pandemic.MethodsMortality cases were from Mar 17 (state of emergency; 0.1 per 100,000), Apr 18 (coronavirus peak; 87.4), Apr 25 (expand testing; 108.7), and May 11 (daily reduced to original; 137.6). Three domains (compositional, contextual, and collective) and 28 county-level predictors of mortality were extracted from American Community Survey, Area Health Resources, US Crime Data, and Religious Data systems for each county. Compositional domain covered socio-demographic characteristics in local areas (e.g., age, sex, race/ethnicity, housing). Contextual domain covered include social and physical opportunities (e.g., health insurance coverage, transportation, mental health providers). Collective domain covered neighborhood safety and religious adherents. Mixed effect regression with the least absolute shrinkage selection operator (LASSO) was used to select the predictors and estimate the parameters after adjusting the time effect and cumulative prevalence of COVID-19.有道词典; 0.1 per 100,000 people详细X;每100000人0.1ResultsNYC and the nearby boroughs (i.e., Bronx, Kings, Manhattan, Queens) had the highest cumulative mortality (231.69 per 100,000 people). Counties far from New York Cities (e.g., Allegany, Cortland, Delaware) had the lowest cumulative mortality. Spatial variation showed counties with larger population density (β=.01, p=.022) and/or higher proportion of people with at least high school education (β=227.24, p=.03) were at risk of higher cumulative mortality in COVID-19.ConclusionUnique spatial clustering mortality risk of COVID-19s was detected, highlighting important but understudied roles of contextual and collective factors. Tailored policy efforts shall be designed to support counties with large population density and high levels of education to prevent the mortality related to COVID-19 infection in NY.Disclosures All Authors: No reported disclosures

On the Application of Entropy Measures with Sliding Window for Intrusion Detection in Automotive In-Vehicle Networks.

The evolution of modern automobiles to higher levels of connectivity and automatism has also increased the need to focus on the mitigation of potential cybersecurity risks. Researchers have proven in recent years that attacks on in-vehicle networks of automotive vehicles are possible and the research community has investigated various cybersecurity mitigation techniques and intrusion detection systems which can be adopted in the automotive sector. In comparison to conventional intrusion detection systems in large fixed networks and ICT infrastructures in general, in-vehicle systems have limited computing capabilities and other constraints related to data transfer and the management of cryptographic systems. In addition, it is important that attacks are detected in a short time-frame as cybersecurity attacks in vehicles can lead to safety hazards. This paper proposes an approach for intrusion detection of cybersecurity attacks in in-vehicle networks, which takes in consideration the constraints listed above. The approach is based on the application of an information entropy-based method based on a sliding window, which is quite efficient from time point of view, it does not require the implementation of complex cryptographic systems and it still provides a very high detection accuracy. Different entropy measures are used in the evaluation: Shannon Entropy, Renyi Entropy, Sample Entropy, Approximate Entropy, Permutation Entropy, Dispersion and Fuzzy Entropy. This paper evaluates the impact of the different hyperparameters present in the definition of entropy measures on a very large public data set of CAN-bus traffic with millions of CAN-bus messages with four different types of attacks: Denial of Service, Fuzzy Attack and two spoofing attacks related to RPM and Gear information. The sliding window approach in combination with entropy measures can detect attacks in a time-efficient way and with great accuracy for specific choices of the hyperparameters and entropy measures.

Artificial Intelligence Solutions for Analysis of X-ray Images.

Artificial intelligence (AI) presents a key opportunity for radiologists to improve quality of care and enhance the value of radiology in patient care and population health. The potential opportunity of AI to aid in triage and interpretation of conventional radiographs (X-ray images) is particularly significant, as radiographs are the most common imaging examinations performed in most radiology departments. Substantial progress has been made in the past few years in the development of AI algorithms for analysis of chest and musculoskeletal (MSK) radiographs, with deep learning now the dominant approach for image analysis. Large public and proprietary image data sets have been compiled and have aided the development of AI algorithms for analysis of radiographs, many of which demonstrate accuracy equivalent to radiologists for specific, focused tasks. This article describes (1) the basis for the development of AI solutions for radiograph analysis, (2) current AI solutions to aid in the triage and interpretation of chest radiographs and MSK radiographs, (3) opportunities for AI to aid in noninterpretive tasks related to radiographs, and (4) considerations for radiology practices selecting AI solutions for radiograph analysis and integrating them into existing IT systems. Although comprehensive AI solutions across modalities have yet to be developed, institutions can begin to select and integrate focused solutions which increase efficiency, increase quality and patient safety, and add value for their patients.

Image Inpainting Detection Based on Multi-task Deep Learning Network

ABSTRACT Image inpainting can effectively repair damaged areas, but it can also be a way of image tampering when it is used to remove meaningful content from an image. Therefore, this paper focuses on the research of inpainting forensics, and proposes a multi-task deep learning method. In order to enhance the learning of texture features, the corresponding local binary pattern channels are added to the input of the network. Furthermore, considering that the multi-task object detection network Mask R-CNN cannot fully utilize the features of all scale feature information during the FPN feature extraction phase, the network in this paper combines Feature Pyramid Networks and back connections to extract more features. This network model can detect not only the images tampered by traditional inpainting methods, but also the images inpainted by deep learning methods. Experimental results on two large public data sets demonstrate the superior performance of the proposed method.

Low Cholesterol Level Linked to Reduced Semantic Fluency Performance and Reduced Gray Matter Volume in the Medial Temporal Lobe.

Hyperlipidemia has been proposed as a risk factor of dementia and cognitive decline. However, the findings of the relationship between cholesterol level and cognitive/brain function have been inconsistent. Here, using a well-controlled sample from the Parkinson’s Progression Markers Initiative (PPMI), we investigated the probable non-linear relationship between plasma total cholesterol (TC) level, gray matter volume (GMv), and cognitive performance in 117 non-demented subjects (mean age, 61.5 ± 8.9 years), including 67 Parkinson’s disease (PD) patients and 50 demographically matched controls. A quadratic relationship between semantic fluency (SF) performance and TC levels was identified. Within the subjects with a desirable TC level (TC < 200 mg/dl), low TC (lTC) levels were associated with reduced SF performance, as well as reduced GMv in three medial temporal regions [including bilateral anterior hippocampus (HIP)]. In contrast, no significant relationship between TC and cognition performance/GMv was found in individuals with a high cholesterol level (i.e., TC ≥ 200 mg/dl). Further region of interest (ROI)-based analysis showed that individuals with TC levels ranging from 100 to 160 mg/dl had the lowest GMv in the medial temporal regions. These findings suggest that low-normal TC level may be associated with reduced cognitive function and brain atrophy in regions implicated in neurodegenerative diseases, adding to a growing body of literature supporting a probable non-linear relationship between cholesterol level and brain health. However, this finding needs to be verified with other large public cohort data that do not include PD patients.

Discovery of Pancreatic Ductal Adenocarcinoma-Related Aberrant Glycosylations: A Multilateral Approach of Lectin Microarray-Based Tissue Glycomic Profiling With Public Transcriptomic Datasets.

Aberrant protein glycosylation is one of the most notable features in cancerous tissues, and thereby glycoproteins with disease-relevant glycosylation alterations are fascinating targets for the development of biomarkers and therapeutic agents. For this purpose, a reliable strategy is needed for the analysis of glycosylation alterations occurring on specific glycoproteins during the progression of cancer. Here, we propose a bilateral approach combining lectin microarray-based tissue glycomic profiling and database-derived transcriptomic datasets. First, lectin microarray was used to perform differential glycomic profiling of crude extracts derived from non-tumor and tumor regions of frozen tissue sections from pancreatic ductal adenocarcinoma (PDAC). This analysis revealed two notable tissue glycome alterations in PDAC samples: increases in sialylated glycans and bisecting N-acetylglucosamine and a decrease in ABO blood group antigens. To examine aberrations in the glycosylation machinery related to these glycomic alterations, we next employed public datasets of gene expression profiles in cancerous and normal pancreases provided by The Cancer Genome Atlas and the Genotype-Tissue Expression projects, respectively. In this analysis, glycosyltransferases responsible for the glycosylation alterations showed aberrant gene expression in the cancerous tissues, consistent with the tissue glycomic profiles. The correlated alterations in glycosyltransferase expression and tissue glycomics were then evaluated by differential glycan profiling of a membrane N-glycoprotein, basigin, expressed in tumor and non-tumor pancreatic cells. The focused differential glycomic profiling for endogenous basigin derived from non-tumor and cancerous regions of PDAC tissue sections demonstrated that PDAC-relevant glycan alterations of basigin closely reflected the notable features in the disease-specific alterations in the tissue glycomes. In conclusion, the present multi-omics strategy using public transcriptomic datasets and experimental glycomic profiling using a tiny amount of clinical specimens successfully demonstrated that basigin is a representative N-glycoprotein that reflects PDAC-related aberrant glycosylations. This study indicates the usefulness of large public data sets such as the gene expression profiles of glycosylation-related genes for evaluation of the highly sensitive tissue glycomic profiling results. This strategy is expected to be useful for the discovery of novel glyco-biomarkers and glyco-therapeutic targets.

Analyzing the Bills-Voting Dynamics and Predicting Corruption-Convictions Among Brazilian Congressmen Through Temporal Networks

In this paper, we propose a network-based technique to analyze bills-voting data comprising the votes of Brazilian congressmen for a period of 28 years. The voting sessions are initially mapped into static networks, where each node represents a congressman and each edge stands for the similarity of votes between a pair of congressmen. Afterwards, the constructed static networks are converted to temporal networks. Our analyses on the temporal networks capture some of the main political changes happened in Brazil during the period of time under consideration. Moreover, we find out that the bills-voting networks can be used to identify convicted politicians, who commit corruption or other financial crimes. Therefore, we propose two conviction prediction methods, one is based on the highest weighted convicted neighbor and the other is based on link prediction techniques. It is a surprise to us that the high accuracy (up to 90% by the link prediction method) on predicting convictions is achieved only through bills-voting data, without taking into account any financial information beforehand. Such a feature makes possible to monitor congressmen just by considering their legal public activities. In this way, our work contributes to the large scale public data study using complex networks.

SAT-307 NORMAL FETAL KIDNEY VOLUME AND INFANT URINALYSIS IN OFFSPRING OF WOMEN TREATED FOR GESTATIONAL DIABETES

The worldwide prevalence of gestational diabetes mellitus (GDM), defined as any degree of glucose intolerance with onset or first recognition during pregnancy, is increasing, and affects up to 20% of pregnancies. Evidence from animal studies indicate that maternal hyperglycemia can be detrimental to renal development and kidney function in exposed offspring. In humans, renal dysfunction and disease has been reported in adults whose mothers had Type 1 or Type 2 diabetes, suggesting that pre-gestational maternal diabetes influences kidney development and the risk of renal dysfunction in adulthood. Despite its high prevalence, few studies have examined kidney development or renal function in offspring of women with GDM. The aim of this study was to evaluate the impact of treated GDM on fetal kidney size and infant renal physiology. A prospective cohort study was conducted between June 2013 and August 2016 in Melbourne, Australia. Participants were recruited at a large public university teaching hospital and clinical data were collected from women without GDM (n=82) and women diagnosed and treated for GDM (n=76). GDM was diagnosed by routine screening at 26-28 weeks gestation. Women with GDM received specialist support for glucose monitoring, diet and exercise in addition to routine antenatal care. One-third of women with GDM also required insulin therapy. Participants underwent an obstetric ultrasound at 32-34 weeks gestation for fetal biometry and fetal kidney volume measurement. Quantitative urinalyses were performed in infant urine samples at 3 months of age for analysis of protein and electrolytes. Maternal age, weight and body mass index were similar in no-GDM and GDM women. Estimated fetal weight and birth weight were also similar in no-GDM and GDM pregnancies. There were no differences in fetal kidney dimensions, total kidney volume, infant urinary albumin or electrolyte levels. Fetal kidney size did not differ by mode of GDM treatment (i.e. diet and exercise alone, or diet and exercise with insulin therapy). There were no associations between maternal glucose levels at screening and fetal total kidney volume or infant urine albumin in multivariable regression analyses. Findings suggest that a limited period of hyperglycemia prior to diagnosis and treatment of GDM in women attending a large metropolitan public hospital in Australia does not alter fetal kidney volume at 32-34 weeks gestation or urinalysis in 3-month-old infants, a favourable finding considering the increasing number of women diagnosed with GDM.

Abstract 1877: The role played by SLUG, an epithelial-mesenchymal transition factor, in invasion and therapeutic resistance of malignant glioma

Abstract Purpose: In malignant gliomas, invasive phenotype and cancer stemness promoting resurgence of residual tumor cells, render treatment very difficult. Hence, identification of epithelial-mesenchymal transition (EMT) factors associated with invasion and stemness of glioma cells is critical. Methods: To address the issue, we investigated several EMT factors in hypermotile U87MG and U251 cells, mouse xenograft model, and human glioma samples. Of several EMT markers, SLUG expression was notably increased at the invasive fronts of gliomas, both in mouse tumor xenografts and human glioma samples. The biological role played by SLUG was investigated using a colony-forming assay after chemotherapy and irradiation, and by employing a neurosphere culture assay. The effect of SLUG on glioma progression was examined in our patient cohort and samples, and compared to large public data from the REMBRANDT and TCGA. Results: Genetic upregulation of SLUG was associated with increased levels of stemness factors and enhanced resistance to radiation and temozolomide. In our cohort, patients exhibiting lower-level SLUG expression evidenced longer progression-free survival (P=0.042). Also, in the REMBRANDT dataset, a group in which SLUG was downregulated exhibited a significant survival benefit (P&amp;lt;0.001). Although paired gliblastoma multiforme (GBM) samples from our patients did not showed significant increase of SLUG expression, increased mRNA levels of SLUG was found in recurred GBM from TCGA (P=0.052), and in TMZ-treated glioma cells and mouse xenograft. Conclusion: SLUG may contribute to glioma progression by controlling invasion at infiltrating margins, associated with increased stemness and therapeutic resistance. Citation Format: Kyung-Hwa Lee, Se-Jeong Oh, Eun-Jung Ahn, Shin Jung, Jae-Hyuk Lee, Kyung-Keun Kim, Hangun Kim, Kyung-Sub Moon. The role played by SLUG, an epithelial-mesenchymal transition factor, in invasion and therapeutic resistance of malignant glioma [abstract]. In: Proceedings of the American Association for Cancer Research Annual Meeting 2019; 2019 Mar 29-Apr 3; Atlanta, GA. Philadelphia (PA): AACR; Cancer Res 2019;79(13 Suppl):Abstract nr 1877.

MultiPLIER: A Transfer Learning Framework for Transcriptomics Reveals Systemic Features of Rare Disease.

Most gene expression datasets generated by individual researchers are too small to fully benefit fromunsupervised machine-learning methods. In the case of rare diseases, there may be too few cases available, even when multiple studies are combined. To address this challenge, we utilize transfer learning to extract coordinated expression patterns and uselearned patterns to analyze small rare disease datasets. We trained a pathway-level information extractor (PLIER) model on a large public data compendium comprising multiple experiments, tissues, and biological conditions and then transferred the model to small datasets in an approach we call MultiPLIER. Models constructed from the public data compendium included features that aligned well to known biological factors and were more comprehensive than those constructed from individual datasets or conditions. When transferred to rare disease datasets, the models describe biological processes related to disease severity more effectively than models trained only on a given dataset.