Normal Dataset Research Articles

Modeling differential car-following behavior under normal and rainy conditions: A memory-based deep learning method with attention mechanism

In order to analyze and learn the difference in car-following behavior between normal and rainy days, we first collect car-following trajectory data of an urban elevated road on normal and rainy days by microwave radar and analyze the differences in speed, relative speed, acceleration, space headway, and time headway among data through statistics. Secondly, owing to the time-series characteristics of car-following data, we use the long short-term memory (LSTM) neural network optimized by attention mechanism (AM) and sparrow search algorithm (SSA) to learn the different car-following behaviors under different weather conditions and build corresponding models (ASL-Normal, ASL-Rain, where ASL stands for AM-SSA-LSTM), respectively. Finally, the simulation test shows that the mean square error (MSE) and reciprocal of time-to-collision (RTTC) of the ASL model are better than those of LSTM and intelligent diver model (IDM), which is closer to the real data. The ASL model can better learn different driving behaviors on normal and rainy days. However, it has a higher sensitivity to weather conditions from cross test on normal and rainy data-sets which need classification training or sample diversification processing. In the car-following platoon simulation, the stability performances of two models are excellent, which can describe the basic characteristics of traffic flow on normal and rainy days. Comparing with ASL-Rain model, the convergence time of ASL-Normal is shorter, reflecting that cautious driving behavior on rainy days will reduce traffic efficiency to a certain extent. However, ASL-Normal model produces a more severe and frequent traffic oscillation within a shorter period because of aggressive driving behavior on normal days.

Investigating Probability Distribution Associated with the Sum of at least Two Independent Positive Integers

This work examined the probabilities associated with the sum of at least two positive integers. Let Y equals the sum of independent positive integers be a random variable. The histograms, probability density function (pdf) and the cumulative probabilities of Y were examined. Besides, the skewness and kurtosis were estimated to be zero (0) and approximately three (3) respectively. The distribution of Y showed clearly bell-shaped curves. Furthermore, the parameters of normal variable Y for at least two positive integers exhibited linear relationship. With these accomplishments it becomes feasible to simulate non-negative normal data sets within certain range of observations available for utilization in further Statistical research intending to mimic real life problems and phenomena. The calculated probabilities, P(Y=yi) and cumulative probabilities, P(Y≤yi) are valuables for inclusion in Standard Statistical Tables both for use by Students and Researchers.

Normal mode observability of radial anisotropy in the Earth’s mantle

SUMMARYObservations of seismic anisotropy provide useful information to infer directions of mantle flow. However, existing global anisotropic tomography models are not consistent, particularly in the lower mantle. Therefore, the interpretation of seismic anisotropy in terms of mantle dynamics and evolution remains difficult. While surface and body waves are commonly used to build radially anisotropic tomography models, they provide heterogeneous data coverage and the radial anisotropy structure retrieved using these data may be biased by the use of imperfect crustal corrections. Normal modes, the free oscillations of the Earth, automatically provide global data coverage and their sensitivity to shear wave (vs) and compressional wave (vp) velocity makes them suitable to study both vs and vp anisotropy in the mantle. In this study, we assess whether current normal mode splitting data have sufficient sensitivity to lower mantle anisotropy to potentially constrain it. We consider the uncertainties in the data and the effect of inaccuracies in crustal thickness corrections and the assumed scaling between vp and vs. We perform forward modelling of normal mode data using six different 3-D global radially anisotropic tomography models to document how strong and widespread anisotropy has to be to be observable in current normal mode data. We find that, on average 50% of the spheroidal and 55% of the toroidal modes investigated show significant sensitivity to vs anisotropy, while roughly 57% of the spheroidal modes also have strong sensitivity to vp anisotropy. Moreover, we find that the normal mode data fit varies substantially for the various anisotropic tomography models considered, with the addition of anisotropy not always improving the data fit. While we find that crustal thickness corrections do not strongly impact modes that are sensitive to the lower mantle, we observe a trade-off between radial anisotropy and vp scaling for these modes. As long as this is taken into consideration, our findings suggest that existing normal mode data sets can provide valuable information on both vs and vp anisotropy in the mantle.

Unsupervised and quantitative intestinal ischemia detection using conditional adversarial network in multimodal optical imaging.

Intraoperative evaluation of bowel perfusion is currently dependent upon subjective assessment. Thus, quantitative and objective methods of bowel viability in intestinal anastomosis are scarce. To address this clinical need, a conditional adversarial network is used to analyze the data from laser speckle contrast imaging (LSCI) paired with a visible-light camera to identify abnormal tissue perfusion regions. Our vision platform was based on a dual-modality bench-top imaging system with red-green-blue (RGB) and dye-free LSCI channels. Swine model studies were conducted to collect data on bowel mesenteric vascular structures with normal/abnormal microvascular perfusion to construct the control or experimental group. Subsequently, a deep-learning model based on a conditional generative adversarial network (cGAN) was utilized to perform dual-modality image alignment and learn the distribution of normal datasets for training. Thereafter, abnormal datasets were fed into the predictive model for testing. Ischemic bowel regions could be detected by monitoring the erroneous reconstruction from the latent space. The main advantage is that it is unsupervised and does not require subjective manual annotations. Compared with the conventional qualitative LSCI technique, it provides well-defined segmentation results for different levels of ischemia. We demonstrated that our model could accurately segment the ischemic intestine images, with a Dice coefficient and accuracy of 90.77% and 93.06%, respectively, in 2560 RGB/LSCI image pairs. The ground truth was labeled by multiple and independent estimations, combining the surgeons' annotations with fastest gradient descent in suspicious areas of vascular images. The total processing time was 0.05s for an image size of . The proposed cGAN can provide pixel-wise and dye-free quantitative analysis of intestinal perfusion, which is an ideal supplement to the traditional LSCI technique. It has potential to help surgeons increase the accuracy of intraoperative diagnosis and improve clinical outcomes of mesenteric ischemia and other gastrointestinal surgeries.

Classification of Normal and Malicious Traffic Based on an Ensemble of Machine Learning for a Vehicle CAN-Network

Connectivity and automation have expanded with the development of autonomous vehicle technology. One of several automotive serial protocols that can be used in a wide range of vehicles is the controller area network (CAN). The growing functionality and connectivity of modern vehicles make them more vulnerable to cyberattacks aimed at vehicular networks. The CAN bus protocol is vulnerable to numerous attacks, as it is lacking security mechanisms by design. It is crucial to design intrusion detection systems (IDS) with high accuracy to detect attacks on the CAN bus. In this paper, we design an effective machine learning-based IDS scheme for binary classification that utilizes eight supervised ML algorithms, along with ensemble classifiers. The scheme achieved a higher effectiveness score in detecting normal and abnormal activities when trained with normal and malicious CAN traffic datasets. Random Forest, Decision Tree, and Xtreme Gradient Boosting classifiers provided the most accurate results. Then we evaluated three ensemble methods, voting, stacking, and bagging, for this classification task. The ensemble classifiers achieved better accuracy than the individual models, since ensemble learning strategies have superior performance through a combination of multiple learning mechanisms. These mechanisms have a varied range of capabilities that improve the prediction reliability while lowering the possibility of classification errors. Our model outperformed the most recent study that used the same dataset, with an accuracy of 0.984.

Detection of Attacks in Network Traffic with the Autoencoder-Based Unsupervised Learning Method

The effects of attacks on network systems and the extent of damages caused by them tend to increase every day. Solutions based on machine learning algorithms have started to be developed in order to develop appropriate defense systems by detecting attacks in a timely and effective manner. This study focuses on detecting abnormal traffic on networks through deep learning algorithms, and a deep autoencoder model architecture that can be used to detect attacks is recommended. To this end, an autoencoder model is first obtained by training the normal dataset without class labels in an unsupervised manner with an autoencoder, and a threshold value is obtained by running this model with small size test data with normal attack observations. The threshold value is calculated as a value that will optimize the model performance. It is observed that supervised learning methods lead to difficulties and cost increases in the detection of cyber-attacks and the labeling process. The threshold value is calculated using only small test data without resorting to labeling in order to overcome these costs and save time, and the incoming up-to-date network traffic information is classified based on this threshold value.

Deep learning-based detection algorithm for brain metastases on black blood imaging

Brain metastases (BM) are the most common intracranial tumors, and their prevalence is increasing. High-resolution black-blood (BB) imaging was used to complement the conventional contrast-enhanced 3D gradient-echo imaging to detect BM. In this study, we propose an efficient deep learning algorithm (DLA) for BM detection in BB imaging with contrast enhancement scans, and assess the efficacy of an automatic detection algorithm for BM. A total of 113 BM participants with 585 metastases were included in the training cohort for five-fold cross-validation. The You Only Look Once (YOLO) V2 network was trained with 3D BB sampling perfection with application-optimized contrasts using different flip angle evolution (SPACE) images to investigate the BM detection. For the observer performance, two board-certified radiologists and two second-year radiology residents detected the BM and recorded the reading time. For the training cohort, the overall performance of the five-fold cross-validation was 87.95%, 24.82%, 19.35%, 14.48, and 18.40 for sensitivity, precision, F1-Score, the false positive average for the BM dataset, and the false positive average for the normal individual dataset, respectively. For the comparison of reading time with and without DLA, the average reading time was reduced by 20.86% in the range of 15.22–25.77%. The proposed method has the potential to detect BM with a high sensitivity and has a limited number of false positives using BB imaging.

Eigen-Entropy: A metric for multivariate sampling decisions

Sampling is a technique to help identify a representative data subset that captures the characteristics of the whole dataset. Most existing sampling algorithms require distribution assumptions of the multivariate data, which may not be available beforehand. This study proposes a new metric called Eigen-Entropy (EE), which is based on information entropy for the multivariate dataset. EE is a model-free metric because it is derived based on eigenvalues extracted from the correlation coefficient matrix without any assumptions on data distributions. We prove that EE measures the composition of the dataset, such as its heterogeneity or homogeneity. As a result, EE can be used to support sampling decisions, such as which samples and how many samples to consider with respect to the application of interest. To demonstrate the utility of the EE metric, two sets of use cases are considered. The first use case focuses on classification problems with an imbalanced dataset, and EE is used to guide the rendering of homogeneous samples from minority classes. Using 10 public datasets, it is demonstrated that two oversampling techniques using the proposed EE method outperform reported methods from the literature in terms of precision, recall, F-measure, and G-mean. In the second experiment, building fault detection is investigated where EE is used to sample heterogeneous data to support fault detection. Historical normal datasets collected from real building systems are used to construct the baselines by EE for 14 test cases, and experimental results indicate that the EE method outperforms benchmark methods in terms of recall. We conclude that EE is a viable metric to support sampling decisions.

Artificial intelligence applied to the classification of eight middle Eocene species of the genus Podocyrtis (polycystine radiolaria)

Abstract. This study evaluates the application of artificial intelligence (AI) to the automatic classification of radiolarians and uses as an example eight distinct morphospecies of the Eocene radiolarian genus Podocyrtis, which are part of three different evolutionary lineages and are useful in biostratigraphy. The samples used in this study were recovered from the equatorial Atlantic (ODP Leg 207) and were supplemented with some samples coming from the North Atlantic and Indian Oceans. To create an automatic classification tool, numerous images of the investigated species were needed to train a MobileNet convolutional neural network entirely coded in Python. Three different datasets were obtained. The first one consists of a mixture of broken and complete specimens, some of which sometimes appear blurry. The second and third datasets were leveled down into two further steps, which excludes broken and blurry specimens while increasing the quality. The convolutional neural network randomly selected 85 % of all specimens for training, while the remaining 15 % were used for validation. The MobileNet architecture had an overall accuracy of about 91 % for all datasets. Three predicational models were thereafter created, which had been trained on each dataset and worked well for classification of Podocyrtis coming from the Indian Ocean (Madingley Rise, ODP Leg 115, Hole 711A) and the western North Atlantic Ocean (New Jersey slope, DSDP Leg 95, Hole 612 and Blake Nose, ODP Leg 171B, Hole 1051A). These samples also provided clearer images since they were mounted with Canada balsam rather than Norland epoxy. In spite of some morphological differences encountered in different parts of the world's oceans and differences in image quality, most species could be correctly classified or at least classified with a neighboring species along a lineage. Classification improved slightly for some species by cropping and/or removing background particles of images which did not segment properly in the image processing. However, depending on cropping or background removal, the best result came from the predictive model trained on the normal stacked dataset consisting of a mixture of broken and complete specimens.

Gradient-based edge detection with skeletonization (GES) segmentation for magnetic resonance optic nerve images

This study proposes a new segmentation method called gradient-based edge detection with skeletonization (GES) for the cross-sectional optic nerve on magnetic resonance (MR) images acquired with T1-weighted fast spoiled gradient-echo (FSPGR) without fat saturation. The raw optic nerve images have very poor resolution with unclear edges. Therefore, the images were first pre-processed with bicubic interpolation to improve the spatial resolution. Then, the proposed GES segmentation was applied to produce a distinct optic nerve image. The edges of the optic nerve were identified by finding the largest gradient changes in signal intensity between the optic nerve region and its surrounding cerebrospinal fluid (CSF). Particle swarm optimization (PSO) and level set method (LSM) segmentations were applied for comparison. Manual segmentation performed by a certified radiologist was used as the ground truth for the evaluation of the computerized segmentation. GES produced a higher mean Dice similarity coefficient (DSC) index of 0.81 ± 0.04 compared to the LSM with a mean DSC index of 0.67 ± 0.17. The bicubic-GES processed optic nerve images were used for the quantitative measurement on ten normal datasets. This study has reported the quantitative values of the longest length of the optic nerve up to the chiasm (37.2 mm) using MR images. The proposed GES segmentation method for the optic nerve will be useful for investigating any optic nerve-related disease that affects the area or volume of the optic nerve.

Nearest neighbor convex hull for health indicator construction

Developing a suitable and monotonous health index (HI) that can be used to represent a whole degradation process is a key step for continuous machine health monitoring during its life cycle. It is expected that the potential HI is able to inform incipient fault moment and then track machine degradation trajectories effectively and monotonically. Previously, nearest neighbor convex hull classification (NNCHC) has been widely applied for fault classification. In this paper, a HI construction methodology for machine life cycle health monitoring based on NNCHC is proposed. Firstly, a normal convex hull is modeled based on normal vibration data to fully characterize machine health conditions. Afterward, two HIs are constructed based on ℓ 1 norm and ℓ 2 norm distances between the normal convex hull and test points. The superiority of the developed approach in this study lies in the flexible and efficient development of a HI for fault progress tracking. Moreover, the only usage of a normal dataset in the proposed methodology is closer to real application.

Augmented data driven self-attention deep learning method for imbalanced fault diagnosis of the HVAC chiller

The chiller fault diagnosis is of great significance to maintain the normal operation of the HVAC system and indoor comfort. Due to the difficulty in collecting the chiller’s fault data, we usually face the data imbalance problem of less fault data and more normal data. To tackle this problem and further improve the fault diagnosis accuracy, this paper proposes a novel augmented data driven self-attention based deep learning model for the chiller fault diagnosis. Firstly, to solve the data imbalance problem, a stable synthetic minority oversampling technique (SSMOTE) is presented to generate the artificial fault data, which are then mixed with the raw fault data to achieve data augmentation. Further, to enhance the diagnosis accuracy, the self-attention mechanism based temporal convolutional network (STCN) is developed to classify normal and fault datasets. The developed STCN is realized by stacking the modified blocks which can dynamically pay attention to different data information through combining the self-attention mechanism and the traditional residual block together. What is more, in the STCN, the skip connection structure is added to the self-attention mechanism to solve the gradient disappearance problem. Finally, detailed experiments and comparisons are performed. Experimental results show that, compared with other data augmentation methods, the SSMOTE can complete a large scale expansion of the minority fault datasets, and its augmented data have better effect on handling the data imbalance problem. Moreover, the skip self-attention mechanism based deep learning model can achieve better diagnosis accuracy compared with some popular deep and shallow models, such as the LSTM, ELM and SVM, etc.

High-resolution transcriptomics informs glial pathology in human temporal lobe epilepsy

The pathophysiology of epilepsy underlies a complex network dysfunction between neurons and glia, the molecular cell type-specific contributions of which remain poorly defined in the human disease. In this study, we validated a method that simultaneously isolates neuronal (NEUN +), astrocyte (PAX6 + NEUN–), and oligodendroglial progenitor (OPC) (OLIG2 + NEUN–) enriched nuclei populations from non-diseased, fresh-frozen human neocortex and then applied it to characterize the distinct transcriptomes of such populations isolated from electrode-mapped temporal lobe epilepsy (TLE) surgical samples. Nuclear RNA-seq confirmed cell type specificity and informed both common and distinct pathways associated with TLE in astrocytes, OPCs, and neurons. Compared to postmortem control, the transcriptome of epilepsy astrocytes showed downregulation of mature astrocyte functions and upregulation of development-related genes. To gain further insight into glial heterogeneity in TLE, we performed single cell transcriptomics (scRNA-seq) on four additional human TLE samples. Analysis of the integrated TLE dataset uncovered a prominent subpopulation of glia that express a hybrid signature of both reactive astrocyte and OPC markers, including many cells with a mixed GFAP + OLIG2 + phenotype. A further integrated analysis of this TLE scRNA-seq dataset and a previously published normal human temporal lobe scRNA-seq dataset confirmed the unique presence of hybrid glia only in TLE. Pseudotime analysis revealed cell transition trajectories stemming from this hybrid population towards both OPCs and reactive astrocytes. Immunofluorescence studies in human TLE samples confirmed the rare presence of GFAP + OLIG2 + glia, including some cells with proliferative activity, and functional analysis of cells isolated directly from these samples disclosed abnormal neurosphere formation in vitro. Overall, cell type-specific isolation of glia from surgical epilepsy samples combined with transcriptomic analyses uncovered abnormal glial subpopulations with de-differentiated phenotype, motivating further studies into the dysfunctional role of reactive glia in temporal lobe epilepsy.

Brain MRI high resolution image creation and segmentation with the new GAN method

Brain magnetic resonance imaging segmentation is a recent and still popular research area. Good and accurate segmentation results play an important role in the diagnosis of cancer or other brain diseases. In this article, a novel Generative Adversarial Network architecture is proposed for brain magnetic resonance imaging segmentation. The proposed method is called SSimDCL (Supervised SimDCL). Four studies were conducted in this article. In the first study, the SSimDCL method on the two-dimensional brain magnetic resonance imaging dataset was compared with the current state-of-art architectures CycleGAN, CUT, FastCUT, DCLGAN, and SimDCL. In the second study, the dataset resolution was improved. In the third study, being measured the efficiency of the newly created dataset. And the SSimDCL is trained for both the dataset with increased resolution and the normal dataset, and the results are obtained. In the fourth study, the results of the SSimDCL and the VolBrain brain magnetic resonance imaging segmentation results, which are widely used today, are included. When VolBrain segmentation and SSimDCL segmentation are compared. The results were compared both visually and metrically. Fréchet Inception Distance (FID), Kernel Inception Distance (KID), Peak Signal to Noise Ratio (PSNR) and Learned Perceptual Image Patch Similarity (LPIPS) were used as measurement metrics. The Jaccard and Dice similarity metrics were also used in the analysis. It was observed that the SSimDCL give satisfactory results in all four studies. This method can be used as an automatic brain MRI image segmentation system.

The effect of Gaussian noise on pneumonia detection on chest radiographs, using convolutional neural networks

IntroductionChest X-rays (CXR) with under-exposure increase image noise and this may affect convolutional neural network (CNN) performance. This study aimed to train and validate CNNs for classifying pneumonia on CXR as normal or pneumonia acquired at different image noise levels. MethodsThe study used the curated and publicly available “Chest X-Ray Pneumonia” dataset of 5856 AP CXR classified into 1583 normal, 4273 viral and bacterial pneumonia cases. Gaussian noise with zero mean was added to the images, at 5 image noise variance levels, corresponding to decreasing exposure. Each noise-level dataset was split into 80% for training, 10% for validation, and 10% for test data and then classified using custom trained sequential CNN architecture. Six classification tasks were developed for five Gaussian noise levels and the original dataset. Sensitivity, specificity, predictive values and accuracy were used as evaluation performance metrics. ResultsCNN evaluation on the different datasets revealed no performance drop from the original dataset to the five datasets with different noise levels. Sensitivity, specificity and accuracy for the normal datasets were 98.7%, 76.1% and 90.2%. For the five Gaussian noise levels the sensitivity, specificity and accuracy ranged from 96.9% to 98.2%, 94.4%–98.7% and 96.8%–97.6%, respectively. A heat map was used for visual explanation of the CNNs. ConclusionThe CNNs sensitivity maintained, and the specificity increased in distinguishing between normal and pneumonia CXR with the introduction of image noise. Implications for practiceNo performance drops of CNNs in distinguishing cases with and without pneumonia CXR with different Gaussian noise levels was observed. This has potential for decreasing radiation dose to patients or maintaining exposure parameters for patients that require additional radiographs.

L-Tetrolet Pattern-Based Sleep Stage Classification Model Using Balanced EEG Datasets.

Background: Sleep stage classification is a crucial process for the diagnosis of sleep or sleep-related diseases. Currently, this process is based on manual electroencephalogram (EEG) analysis, which is resource-intensive and error-prone. Various machine learning models have been recommended to standardize and automate the analysis process to address these problems. Materials and methods: The well-known cyclic alternating pattern (CAP) sleep dataset is used to train and test an L-tetrolet pattern-based sleep stage classification model in this research. By using this dataset, the following three cases are created, and they are: Insomnia, Normal, and Fused cases. For each of these cases, the machine learning model is tasked with identifying six sleep stages. The model is structured in terms of feature generation, feature selection, and classification. Feature generation is established with a new L-tetrolet (Tetris letter) function and multiple pooling decomposition for level creation. We fuse ReliefF and iterative neighborhood component analysis (INCA) feature selection using a threshold value. The hybrid and iterative feature selectors are named threshold selection-based ReliefF and INCA (TSRFINCA). The selected features are classified using a cubic support vector machine. Results: The presented L-tetrolet pattern and TSRFINCA-based sleep stage classification model yield 95.43%, 91.05%, and 92.31% accuracies for Insomnia, Normal dataset, and Fused cases, respectively. Conclusion: The recommended L-tetrolet pattern and TSRFINCA-based model push the envelope of current knowledge engineering by accurately classifying sleep stages even in the presence of sleep disorders.

Intelligent based hybrid renewable energy resources forecasting and real time power demand management system for resilient energy systems.

The rapid growth of hybrid renewable Distributed Energy Resources (DERs) generation possess various challenges with inaccurate forecast models in stochastic power systems. The prime objective of this research is to maximum utilization of scheduled power from hybrid renewable based DERs to maintain the load-demand profile with reduce distributed grid burden. The proposed mixed input-based cascaded artificial neural network is realized for the prediction of a short-term based hourly solar irradiance and wind speed. The testing approach is performed through a historical hourly dataset of the proposed site. Further, the normalized data sets are divided into hourly-based samples for validating the load demand power with respect to the variation in metrological data. In this paper, Adaptive Neuro-Fuzzy Inference System (ANFIS) model is simulated for short-term power demand prediction. This adaptive methodology is an effective approach for load-demand management which is based on cross-entropy. It also confirmed that during testing, the forecasting mean error and cross-entropy are less than 5% under a specific time slap of an individual day. The regression analysis is performed through the time series fitting simulation tool at different time horizons. The performance evaluation of the designed model is compared with the multi-layer perceptron model. Simulation results display the proposed mixed input-based cascaded system has enhanced accuracy and optimal performance than the multi-output correlated perceptron model.

Combination of Astrogliosis and Phosphorylated Tau for the Preclinical Diagnosis of Alzheimer Disease Using 3-Dimensional Stereotactic Surface Projection Images With 18 F-THK5351.

The Alzheimer disease (AD) brain is characterized microscopically by the presence of extracellular amyloid plaques and intraneuronal neurofibrillary tangles consisting of phosphorylated tau aggregations. 18 F-THK5351 is a first-generation PET tau tracer that also binds to monoamine oxidase B, which represents astrogliosis, and is useful to evaluate some non-AD neurodegenerative disorders. We examined the utility of 18 F-THK5351 in preclinical AD using 3-dimensional stereotactic surface projection images optimized for its pathological accumulation by comparison with a normal dataset. By using this 18 F-THK5351 3-dimensional stereotactic surface projection procedure, which can evaluate phosphorylated tau and neuroinflammation, we could diagnose preclinical AD effectively.

IfCNV: A novel isolation-forest-based package to detect copy-number variations from various targeted NGS datasets

Copy-number variations (CNVs) are an essential component of genetic variation distributed across large parts of the human genome. CNV detection from next-generation sequencing data and artificial intelligence algorithms have progressed in recent years. However, only a few tools have taken advantage of machine-learning algorithms for CNV detection, and none propose using artificial intelligence to automatically detect probable CNV-positive samples. The most developed approach is to use a reference or normal dataset to compare with the samples of interest, and it is well known that selecting appropriate normal samples represents a challenging task that dramatically influences the precision of results in all CNV-detecting tools. With careful consideration of these issues, we propose here ifCNV, a new software based on isolation forests that creates its own reference, available in R and python with customizable parameters. ifCNV combines artificial intelligence using two isolation forests and a comprehensive scoring method to faithfully detect CNVs among various samples. It was validated using targeted next-generation sequencing (NGS) datasets from diverse origins (capture and amplicon, germline and somatic), and it exhibits high sensitivity, specificity, and accuracy. ifCNV is a publicly available open-source software (https://github.com/SimCab-CHU/ifCNV) that allows the detection of CNVs in many clinical situations.

Abstract A009: Proteogenomic prioritization of immunotherapeutic targets in rhabdomyosarcoma nominate MEGF10 for preclinical development

Abstract Background: Rhabdomyosarcoma (RMS) has a high unmet need in terms of precision therapy development as there are currently no approved immunotherapies or targeted therapies, and few in the developmental pipeline. Here we sought to identify cell surface oncoproteins as a target for novel RMS-directed immunotherapies. Methods: We first performed plasma membrane enrichment followed by mass spectrometry to define the cell surface landscape of 7 fusion(+) and 14 fusion(-) RMS patient-derived xenograft (PDX) models. The surfaceome data was filtered to only “high confidence” surface proteins by querying protein localization databases, Compartments (https://compartments.jensenlab.org/) and CIRFESS (https://gundrylab.shinyapps.io/cirfess/). We then developed a prioritization algorithm that uses a rank-product approach to score surface proteins. The input to the algorithm is a matrix that integrates multiple datasets to score the surface proteins based on their suitability to be an optimal immunotherapeutic target. In addition to the surfaceome data generated here, we also integrated matched RNA-sequencing data from eleven of the RMS PDX models, RNA-sequencing data from GTEx (n=15,253) and a recently developed normal tissue proteomics dataset (n=201) [Jiang. Cell. 2020], a list from Gene Ontology that included genes involved in muscle development pathways, and the gene dependency list for RMS in DepMap (https://depmap.org/portal/). Results: A total of 913 and 937 high confidence surface proteins were annotated from the mass spectrometry data for fusion(+) and fusion(−) samples, respectively. A dendrogram separated the surface protein profiles into two clusters based on fusion(+) and fusion(−) RMS subtypes, thus the algorithm was run separately on each subtype. Within the top 50% of prioritized targets, 88% and 86% of the targets overlapped and 12% and 14% were identified exclusively in fusion(+) and fusion(−) subtypes, respectively. ALK, a previously putative protein marker in fusion(+) RMS, scored in the top 10% of the fusion(+) targets based on the algorithm, and surprisingly we saw abundant ALK expression in 6/14 fusion(−) PDX samples. MEGF10, a novel target, was ranked as the top target for both fusion(+) and fusion(−) RMS. MEGF10 plays a role in cell adhesion, motility, and proliferation. It scored as a significant dependency in DepMap for RMS (p-value=0.0002). Based on RNA-sequencing and proteomics, MEGF10 shows no expression in most healthy tissues surveyed, with several orders of magnitude lower expression detected in RNASeq in muscle and brain tissue, but not in the proteomic datasets. Conclusion: Here, we defined the surfaceome of RMS, and found substantial overlap in surface proteins between fusion(+) and fusion(−) RMS subtypes. We validated previous observations that ALK is expressed in RMS, here verifying that the protein is expressed on the plasma membrane. MEGF10 appears to be a strong novel candidate target for RMS immunotherapies, and ongoing work to validate our proteogenomic findings will be reported. Citation Format: Rawan Shraim, Amber K. Weiner, Karina L. Conkrite, Alexander B. Radaoui, John M. Maris, Yael P. Mosse, Sharon J. Diskin, Ahmet Sacan, Benjamin A. Garcia, Peter J. Houghton, Raushan T. Kurmasheva, Poul Sorensen, Gregg B. Morin, Brian Mooney. Proteogenomic prioritization of immunotherapeutic targets in rhabdomyosarcoma nominate MEGF10 for preclinical development [abstract]. In: Proceedings of the AACR Special Conference: Sarcomas; 2022 May 9-12; Montreal, QC, Canada. Philadelphia (PA): AACR; Clin Cancer Res 2022;28(18_Suppl):Abstract nr A009.