Mahalanobis Distance Method Research Articles

Predictive Model for Diagnosis of Gestational Diabetes in the Kurdistan Region by a Combination of Clustering and Classification Algorithms: An Ensemble Approach

Gestational diabetes is a type of high blood sugar that develops during pregnancy. It can occur at any stage of pregnancy and cause problems for both the mother and the baby, during and after birth. The risks can be reduced if they are early detected and managed, especially in areas where only periodic tests of pregnant women are available. Intelligent systems designed by machine learning algorithms are remodelling all fields of our lives, including the healthcare system. This study proposes a combined prediction model to diagnose gestational diabetes. The dataset was obtained from the Kurdistan region laboratories, which collected information from pregnant women with and without diabetes. The suggested model uses the clustering KMeans technique for data reduction and the elbow method to find the optimal k value and the Mahalanobis distance method to find more related cluster to new samples, and the classification methods such as decision tree, random forest, SVM, KNN, logistic regression, and Naïve Bayes are used for prediction. The results showed that using a mix of KMeans clustering, elbow method, Mahalanobis distance, and ensemble technique significantly improves prediction accuracy.

OUTLIER DETECTION ON HIGH DIMENSIONAL DATA USING MINIMUM VECTOR VARIANCE (MVV)

High-dimensional data can occur in actual cases where the variable p is larger than the number of observations n. The problem that often occurs when adding data dimensions indicates that the data points will approach an outlier. Outliers are part of observations that do not follow the data distribution pattern and are located far from the data center. The existence of outliers needs to be detected because it can lead to deviations from the analysis results. One of the methods used to detect outliers is the Mahalanobis distance. To obtain a robust Mahalanobis distance, the Minimum Vector Variance (MVV) method is used. This study will compare the MVV method with the classical Mahalanobis distance method in detecting outliers in non-invasive blood glucose level data, both at p>n and n>p. The test results show that the MVV method is better for n>p. MVV shows more effective results in identifying the minimum data group and outlier data points than the classical method.

Comparing outcomes from tailored meta-analysis with outcomes from a setting specific test accuracy study using routine data of faecal calprotectin testing for inflammatory bowel disease

BackgroundMeta-analyses of test accuracy studies may provide estimates that are highly improbable in clinical practice. Tailored meta-analysis produces plausible estimates for the accuracy of a test within a specific setting by tailoring the selection of included studies compatible with a specific setting using information from the target setting. The aim of this study was to validate the tailored meta-analysis approach by comparing outcomes from tailored meta-analysis with outcomes from a setting specific test accuracy study.MethodsA retrospective cohort study of primary care electronic health records provided setting-specific data on the test positive rate and disease prevalence. This was used to tailor the study selection from a review of faecal calprotectin testing for inflammatory bowel disease for meta-analysis using the binomial method and the Mahalanobis distance method. Tailored estimates were compared to estimates from a study of test accuracy in primary care using the same routine dataset.ResultsTailoring resulted in the inclusion of 3/14 (binomial method) and 9/14 (Mahalanobis distance method) studies in meta-analysis. Sensitivity and specificity from tailored meta-analysis using the binomial method were 0.87 (95% CI 0.77 to 0.94) and 0.65 (95% CI 0.60 to 0.69) and 0.98 (95% CI 0.83 to 0.999) and 0.68 (95% CI 0.65 to 0.71), respectively using the Mahalanobis distance method. The corresponding estimates for the conventional meta-analysis were 0.94 (95% CI 0.90 to 0.97) and 0.67 (95% CI 0.57 to 0.76) and for the FC test accuracy study of primary care data 0.93 (95%CI 0.89 to 0.96) and 0.61 (95% CI 0.6 to 0.63) to detect IBD at a threshold of 50 μg/g. Although the binomial method produced a plausible estimate, the tailored estimates of sensitivity and specificity were not closer to the primary study estimates than the estimates from conventional meta-analysis including all 14 studies.ConclusionsTailored meta-analysis does not always produce estimates of sensitivity and specificity that lie closer to the estimates derived from a primary study in the setting in question. Potentially, tailored meta-analysis may be improved using a constrained model approach and this requires further investigation.

Minimum regularized covariance determinant and principal component analysis-based method for the identification of high leverage points in high dimensional sparse data

ABSTRACT The main aim of this paper is to propose a novel method (RMD-MRCD-PCA) of identification of High Leverage Points (HLPs) in high-dimensional sparse data. It is to address the weakness of the Robust Mahalanobis Distance (RMD) method which is based on the Minimum Regularized Covariance Determinant (RMD-MRCD), which indicates a decrease in its performance as the number of independent variables (p) increases. The RMD-MRCD-PCA is developed by incorporating the Principal Component Analysis (PCA) in the MRCD algorithm whereby this robust approach shrinks the covariance matrix to make it invertible and thus, can be employed to compute the RMD for high dimensional data. A simulation study and two real data sets are used to illustrate the merit of our proposed method compared to the RMD-MRCD and Robust PCA (ROBPCA) methods. Findings show that the performance of the RMD-MRCD is similar to the performance of the RMD-MRCD-PCA for p close to 200. However, its performance tends to decrease when the number of p is more than 200 and worsens at p equals 700 and larger. On the other hand, the ROBPCA is not effective for less than 20% contamination as it suffers from serious swamping problems.

Ontology based sentiment analysis for fake review detection

Majority of customers and manufacturers who tend to purchase and trade via e-commerce websites primarily rely on reviews before making purchasing decisions and product improvements. Deceptive reviewers consider this opportunity to write fake reviews to mislead customers and manufacturers. This calls for the necessity of identifying fake reviews before making them available for decision making. Accordingly, this research focuses on a fake review detection method that incorporates review-related features including linguistic features, Part-of-Speech (POS) features, and sentiment analysis features. A domain feature ontology is used in the feature-level sentiment analysis and all the review-related features are extracted and integrated into the ontology. The fake review detection is enhanced through a rule-based classifier by inferencing the ontology. Due to the lack of a labeled dataset for model training, the Mahalanobis distance method was used to detect outliers from an unlabeled dataset where the outliers were selected as fake reviews for model training. The performance measures of the rule-based classifier were improved by integrating linguistic features, POS features, and sentiment analysis features, in spite of considering them separately.

A tailored use of the mahalanobis distance matching for causal effects estimation: A simulation study

Selection bias is a fundamental reason why the estimation of treatment effects in observational studies is not as straightforward as in well-designed experiments. We introduce an improved and generalized matching method based on redefining how the Mahalanobis distance can be used for matching to reduce covariate imbalance and improve the efficiency of treatment effect estimates. The motivation for this new proposal is based on the inability of the previously proposed covariate balancing rank-based Mahalanobis distance (CBRMD) method in explicitly creating fair play in its weight allocation. Our newly proposed approach has a particularly interesting parameter that provides insight into the behaviour of less variable and less aberrant weights that can maximize covariate balance. We illustrated the proposed technique by analyzing simulated data and two real-life datasets. In addition, a web-based Shiny application written in R statistical language was developed and deployed online to implement the proposed technique. Our new proposal is a promising alternative tool for reducing selection bias in the estimation of causal effects.

A two-dimensional sample screening method based on data quality and variable correlation

The selection of a training set is the key to determining the quality of the model. In the spectrum analysis, due to various interference factors, the quality of the collected spectral data of some samples has a serious deviation. If directly used in modeling, it will introduce bias to the establishment of the model. Therefore, to get the most representative samples, it is necessary to select samples before establishing the model. This paper proposes a two-dimensional sample selection (TDSS) method, which selects samples from two angles of spectral data quality and variable correlation. This method and Mahalanobis distance method were respectively applied to dynamic spectrum (DS) data to screen samples. The samples screened by the two methods were used for modeling. Finally, establish partial least squares (PLS) linear regression model with a quadratic nonlinear correction method to predict the target components. The experimental results show that the sample screening method significantly improved the accuracy and prediction performance of the model, and it is better than the Mahalanobis distance method. In the prediction of triglyceride and total cholesterol, the correlation coefficient can reach above 0.82. The experimental results fully prove the effectiveness of the sample selection method in this paper, and it has a remarkable effect on improving the accuracy and robustness of the model. This paper provides a new way for sample selection of modeling set in spectral analysis of complex solutions.

A method of predicting oil and gas resource spatial distribution based on Bayesian network and its application

Abstract The spatial distribution prediction of hydrocarbon resource is essential to reduce exploration risks and improve investment returns. Determining the location of the spatial distribution of oil and gas resources is a complex and uncertain problem. This paper systematically analyzes oil and gas exploration risk assessment and resource spatial distribution prediction technology, and proposes a method of predicting hydrocarbon spatial distribution based on averaged one-dependence estimators(AODE). Firstly, the transformation process of petroleum geological problem to mathematical model is described. Then, the classification principle and decision rule of AODE model are expounded. Finally, a case study of Sangonghe Formation in the hinterland of Junggar Basin in China is given to further illustrate the proposed method and work flow. The case prediction results not only reveal that the accuracy of the AODE model can reach 85.2% on the data set of 203 exploratory wells, which is higher than state-of-the-art methods (such as tree augmented Bayesian network method and the Mahalanobis distance method), but also point out the areas with high probability of remaining hydrocarbon resources in Sangonghe Formation, which provides decision-making basis for the next exploration. The application results show that the AODE model can effectively predict the spatial distribution of oil and gas and visualize the risk of geological exploration, thereby optimizing drilling strategy and increasing economic benefits.

Genetic Divergence and Selection of Common Bean Genotypes with High Agronomic Performance in Mozambique

The deep knowledge of genetic diversity in access, in the breeding program, and the contribution of genetic factors as well as the phenotypic expression is of extreme importance for the development of new cultivars and maintenance of the program. This research aimed to study the genetic diversity and selection of the best common bean genotypes with high agronomic performance in the Agrarian Station of Sussundenga-Mozambique. Twenty-two common bean genotypes were evaluated in a randomized block design with three repetitions in the 2021/2022 agricultural season. The agronomic characteristics evaluated were: the number of days to flowering, plant height, plant architecture, number of pods per plant, the weight of 100 seeds, and grain yield. The data obtained were submitted to individual variance analysis, followed by the Scott-Knott grouping test. For the effect of genetic divergence between the genotypes, multivariate analysis was used based on the generalized Mahalanobis distance and Tocher optimization method, UPGMA, and canonical variables for the projection of distance in the 2D plane. It was also verified by the Cophenetic Correlation Coefficient. The Singh criterion (1981) was used, indicating the characteristics Number of pods per plant (42.81%), the weight of 100 seeds (19.59%), and Plant height (16.08%) as the most important for the genetic divergence of bean genotypes.

A Critical Overview of FDA and EMA Statistical Methods to Compare In Vitro Drug Dissolution Profiles of Pharmaceutical Products.

A drug dissolution profile is one of the most critical dosage form characteristics with immediate and controlled drug release. Comparing the dissolution profiles of different pharmaceutical products plays a key role before starting the bioequivalence or stability studies. General recommendations for dissolution profile comparison are mentioned by the EMA and FDA guidelines. However, neither the EMA nor the FDA provides unambiguous instructions for comparing the dissolution curves, except for calculating the similarity factor f2. In agreement with the EMA and FDA strategy for comparing the dissolution profiles, this manuscript provides an overview of suitable statistical methods (CI derivation for f2 based on bootstrap, CI derivation for the difference between reference and test samples, Mahalanobis distance, model-dependent approach and maximum deviation method), their procedures and limitations. However, usage of statistical approaches for the above-described methods can be met with difficulties, especially when combined with the requirement of practice for robust and straightforward techniques for data evaluation. Therefore, the bootstrap to derive the CI for f2 or CI derivation for the difference between reference and test samples was selected as the method of choice.

Classification of chronic myeloid leukemia neutrophils by hyperspectral imaging using Euclidean and Mahalanobis distances

Chronic Myeloid Leukemia (CML) is a type of blood cancer which needs to be diagnosed in early stages to facilitate effective treatment. This necessitates quick, error free and automated diagnostic techniques. In this study, hyperspectral images have been analyzed using statistical distances to classify neutrophils from CML versus healthy blood samples. The statistical distances were used in multidimensional space offered by hyperspectral images. For computational efficiency, principal component analysis was used to achieve dimensionality reduction. The Euclidean distance method, and Mahalanobis distance method which compensates the variance of the target data distribution were used to classify CML neutrophils. The effectiveness of the proposed methods were tested and compared using experimental results. The Euclidean distance was found to be superior when it came to sensitivity in detecting CML neutrophils whereas the Mahalanobis distance was better at detecting healthy neutrophils and distinguishing CML neutrophils from healthy neutrophils.

TRK Fusion Cancer: Patient Characteristics and Survival Analysis in the Real-World Setting.

BackgroundNeurotrophic tyrosine receptor kinase (NTRK) gene fusions are oncogenic drivers in various tumor types. While NTRK gene fusions are predictive of benefit from tropomyosin receptor kinase inhibitors regardless of tumor type, the prognostic significance of NTRK gene fusions in a pan-tumor setting remains unclear.ObjectiveThis study evaluated the characteristics and prognosis of tropomyosin receptor kinase fusion cancer in the real-world setting.Patients and MethodsThis retrospective study used a de-identified clinico-genomic database and included patients with cancer who had comprehensive genomic profiling between January 2011 and July 2018. Patients were classified as having cancer with NTRK gene fusions or NTRK wild-type genes. Patients were matched with a 1:4 ratio (NTRK fusion:NTRK wild-type) using the Mahalanobis distance method on demographic and clinical characteristics, including age and Eastern Cooperative Oncology Group performance status. Descriptive analysis of clinical and molecular characteristics was conducted. Kaplan–Meier estimator and Cox regression were used for overall survival analysis.ResultsMedian overall survival was 12.5 months (95% confidence interval 9.5–not estimable) and 16.5 months (95% confidence interval 12.5–22.5) in the NTRK gene fusion (n = 27) and NTRK wild-type cohorts (n = 107), respectively (hazard ratio 1.44; 95% confidence interval 0.61–3.37; p = 0.648). Co-occurrence of select targetable biomarkers including ALK, BRAF, ERBB2, EGFR, ROS1, and KRAS was lower in cancers with NTRK gene fusions than in NTRK wild-type cancers.ConclusionsAlthough the hazard ratio for overall survival suggested a higher risk of death for patients with NTRK gene fusions, the difference was not statistically significant. Co-occurrence of NTRK gene fusions and other actionable biomarkers was uncommon.Supplementary InformationThe online version contains supplementary material available at 10.1007/s11523-021-00815-4.

마할라노비스 거리법을 이용한 절단 품질에 대한 연구

Social interest in the 4th industry, intelligent factories, and smart manufacturing is continually growing along with the core technologies like big data and artificial intelligence, which can generate meaningful information by collecting and accumulating sensor data. Demand for industrial automation equipment is increasing worldwide due to the efforts needed to modernize manufacturing facilities, reduce automation and cycle time, and improve quality. Currently, the majority of research is focused on the development of automation facilities and improving productivity. The research on the contents of real-time data considering the characteristics of the cutting machine plasma machine is insufficient. In this study, based on the current data measured according to cutting current and cutting speed, a reference value for cutting quality is presented and the optimal process parameter has been selected. A model for predicting cutting quality by introducing the Mahalanobis Distance Method is presented. An attempt has been made to derive selection and optimal cutting process variables. Based on the predictive model, threshold values were specified and used in real-time data to consider the correlations between multivariate variables and evaluate the degree of scattering around the average of specific values of each variable. Also, process parameters suitable for surface roughness were calculated.

Adapting the QUEFTS model to predict attainable yields when training data are characterized by imperfect management

Abstract Understanding yield responses to nutrient application is a key input for extension advice and strategic agricultural investments in developing countries. A commonly used model for yield responses to nutrient inputs in tropical smallholder farming systems is QUEFTS (QUantitative Evaluation of the Fertility of Tropical Soils). While QUEFTS has a strong conceptual foundation, a key assumption is that nutrients are the only limiting factors. One implication of this is the required assumption of ‘perfect management’. This may be problematic in the application of QUEFTS in smallholder farming systems with a wide variety of yield limiting factors. In a previous study, QUEFTS was calibrated using farm trials in two major maize production zones in Nigeria. To reduce observed variability in correlations between estimated soil nutrient (N, P, K) supply and soil parameters (e.g. soil organic carbon, soil pH; step 1 of QUEFTS) a Mahalanobis distance method was used to remove data points not adhering to expected correlations. In this study, we assessed an alternative approach: can the QUEFTS model be adapted to fit smallholder farming systems and associated variation in management? Using 676 observations from the same nutrient omission trials in two major maize production zones in Nigeria, we compare a standard linear regression approach with a quantile regression approach to calibrate QUEFTS. We find that under the standard linear regression approach, there is a poor relation between predicted and observed yields. Using quantile regression, however, QUEFTS performed better at predicting attainable yields – defined as the 90th percentile of observed yields – under a wide variety of production conditions. Our results indicate that using quantile regression as a way to predict attainable yields, is a useful alternative implementation of QUEFTS in smallholder farming systems with high variability in management and other characteristics.

Anomaly Detection in Airborne Fourier Transform Thermal Infrared Spectrometer Images Based on Emissivity and a Segmented Low-Rank Prior

Although hyperspectral anomaly detection is commonly conducted in the visible, near-infrared, and shortwave infrared spectral regions, there has been less research on hyperspectral anomaly detection in the longwave infrared (LWIR) hyperspectral region. The radiance of thermal infrared hyperspectral imagery is determined by the temperature and emissivity. To avoid the detection uncertainty caused by the single factor of temperature, emissivity can be introduced to detect anomalies. However, in the emissivity domain, the spectral contrast and signal-to-noise ratio (SNR) are low, which makes it difficult to separate the anomalies from the background. In this paper, an anomaly detection method combining emissivity and a segmented low-rank prior (EaSLRP) is proposed for use with thermal infrared hyperspectral imagery. The EaSLRP method is divided into three parts—1) temperature/emissivity retrieval, 2) extraction of the thermal infrared hyperspectral background information, and 3) Mahalanobis distance detection. A homogeneous region generation method is also proposed to solve the problem of the complex global background leading to inaccurate background estimation. The GoDec method is used for matrix decomposition and background information extraction and to remove some of the noise. The proposed Mahalanobis distance detector then uses the background component and original image for anomaly detection, while highlighting the spectral difference between the anomalies and background. This method can also suppress the influence of noise, to some extent. The experimental results obtained with airborne Fourier transform thermal infrared spectrometer hyperspectral images demonstrate that the EaSLRP method is effective when compared with the Reed–Xiaoli detector (RXD), the segmented RX detector (SegRX), the low-rank and sparse representation-based detector (LRASR), the low-rank and sparse matrix decomposition (LRaSMD)-based Mahalanobis distance method (LSMAD), and the locally enhanced low-rank prior method (LELRP-AD).

Identification of plant species in an alpine steppe of Northern Tibet using close-range hyperspectral imagery

The identification of plant species in alpine steppes of Northern Tibet is of great significance for revealing community structures and for monitoring vegetation degradation and restoration from remote sensing images. Plants in the alpine steppe of Northern Tibet are short, sparse, and highly heterogeneous in spatial distribution. This peculiarity makes the plant species identification here much more difficult than the identification of plants with high spatial homogeneity. We aimed to explore the potential of close-range hyperspectral imaging for plant species identification in alpine steppe under field conditions. Specifically, we assessed which spectral bands are effective and which classification methods are suitable for plant species identification. A close-range hyperspectral image of grassland in Nagqu, Tibet were acquired in August 2018. Four methods, including derivatives, continuum removal, spectral indices, and principal components were used to enhance the differences in spectral characteristics between plant species. Then, two band selection methods, including Mahalanobis distance and variable importance evaluations based on a random forest (RF) were used to reduce dimensionality and select indicators beneficial for identifying grass species. Four datasets were constructed based on those indicators and were used as the input data for four classifiers, support vector machine (SVM), RF, artificial neural network (ANN), and spectral angle mapper (SAM). We found that (1) bands selected using Mahalanobis distance and variable importance evaluation method showed that the red bands, red edge bands, and spectral indices were important for plant species identification; (2) among the four classifiers, the ANN classifier had the highest overall classification accuracy on Dataset 3 of the reflectance images, which was 94.73%, and the Kappa coefficient was 0.93; (3) the machine learning algorithms RF and ANN performed well for identifying plant species, with an overall accuracy more than 91.59% and kappa coefficient above 0.89. These results suggest that close-range hyperspectral image and machine-learning classifiers, such as RF and ANN, can be used to effectively identify plant species in alpine steppe.

Micrometeoroid and Orbital Debris Impact Detection and Location Based on FBG Sensor Network Using Combined Artificial Neural Network and Mahalanobis Distance Method

Micrometeoroid and orbital debris (MMOD) can cause serious impact damage to long-term on-orbit spacecraft. MMOD impact detection and location are essential to improve the safety and reliability of the on-orbit spacecraft. In this article, a combined artificial neural network (ANN) and Mahalanobis distance method for MMOD impact detection and location based on fiber Bragg grating (FBG) sensor network is proposed. The inputs of the proposed ANN are the Mahalanobis distances determined by the wavelength changes of the FBG sensors at impact points and reference points. A finite element model is developed to explore the relationship between the Mahalanobis distance and the actual distance, which proved to be nonlinear based on the finite element simulation results of MMOD impacts. Four FBG sensors were installed on the back of the test board and sampled at a frequency of 100 Hz using a four-channel FBG demodulator. A total of 360 impact experiments were carried out on this experiment system. The average location error of the proposed method is 0.89 cm, which is 39% and 31% lower compared with the Mahalanobis distance discriminant analysis method and the ANN method, respectively. The combined ANN and Mahalanobis distance method reduces the location errors and has higher accuracy of position prediction.

Welding Process Quality Improvement with Machine Learning Techniques

This paper presents a methodology for selecting design parameters with influence in quality product, using data analysis and machine learning techniques applied to a manufacturing problem where execution results can provide data for design improvement. Techniques are developed and used to extract the maximum information from existing process and quality data, and therefore to get feedback at design stages regarding how parameters variations affect the product performance when certain quality tests are applied. An algorithm is developed to sample multidimensional points in order to get the most representative subset, and various techniques are applied for data cleaning and outliers detection, such as Mahalanobis distance, Cook distance and HDOutliers method. Finally, inductive learning techniques, such as decision trees and association rules, are implemented to get conclusions about the relationships of the different variables. This paper finally shows this methodology in an automotive welding scenario.

Infrared Thermography-Based Fault Diagnosis of Induction Motor Bearings Using Machine Learning

Bearing is one of the most crucial parts in induction motor (IM) as a result there is a constant call for effective diagnosis of bearing faults for reliable operation. Infrared thermography (IRT) is appreciably used as a non-destructive and non-contact method to detect the bearing defects in a rotary machine. However, its performance is limited by insignificant information and string noise present in the infrared thermal image. To address this issue, an emergent two dimensional discrete wavelet transform (2D-DWT) based IRT method has been proposed in this article for diagnosing the different bearing faults in IM, namely, inner and outer race defects, and lack of lubrication. The dimensionality of the extracted features was reduced using principal component analysis (PCA) and thereafter the selected features were ranked in the order of most relevant features using the Mahalanobis distance (MD) method to achieve the optimal feature set. Finally these selected features have been passed to the complex decision tree (CDT), linear discriminant analysis (LDA) and support vector machine (SVM) for fault classification and performance evaluation. The classification results reveal that the SVM outperformed CDT and LDA. The proposed strategy can be used for self-adaptive recognition of bearing faults in IM which helps to avoid the unplanned and unwanted system shutdowns due to the bearing failure.

Abstract 09: Cancers with NTRK gene fusions: Molecular characteristics and prognosis

Abstract Introduction: The neurotrophic tyrosine receptor kinase (NTRK) genes NTRK1, 2, and 3 have emerged as oncogenic drivers, and their resultant tropomyosin receptor kinase (TRK) proteins are targets for precision medicines. Larotrectinib, a selective TRK inhibitor, and entrectinib, a multikinase inhibitor, have been approved for patients (pts) with TRK fusion cancer. This study was conducted to evaluate the molecular characteristics and prognosis of TRK fusion cancer in a real-world setting. Methods: This is a retrospective study using the de-identified Flatiron Health-Foundation Medicine Clinico-Genomic Database (CGDB) that includes pts with a diagnosis of cancer who had comprehensive genomic profiling (CGP) between Jan 2011 and Jul 2018 (Version Nov. 2018). Pts were classified based on the status of their NTRK gene: cohort 1 included pts with NTRK gene fusions and cohort 2 included pts with wild-type NTRK genes (WT-TRK). Within each tumor type, matching was conducted between the two cohorts using the Mahalanobis distance method with a ratio of 1:4 (cohort 1: cohort 2) on demographic and clinical characteristics including age, albumin, antineoplastic use, ECOG performance status, practice type, and year of CGP report. Descriptive analyses were provided for patient and molecular characteristics. Kaplan-Meier method and Cox regression were used for overall survival (OS) analysis with the date of the CGP report as the index date. Results: Of 33,398 pts in the CGDB, 29 pts across 18 histologies had an NTRK gene fusion and 12,456 pts had the same histologies with WT-TRK. In pts with NTRK gene fusions, NTRK1 was involved in 18 (62.1%), NTRK2 in 3 (10.3%), and NTRK3 in 8 (27.6%) of the observed fusion events. Co-occurrence of specific biomarkers (ALK, BRAF, ERBB2, EGFR, ROS1, KRAS) was rare, and lower in pts with NTRK gene fusions than in WT-TRK pts; however, high TMB and MSI were more common in pts with NTRK gene fusions than in WT-TRK pts (TMB high 20.7% vs. 2.8% and MSI high 10.3% vs. 0.9%, respectively). 27 pts from cohort 1 were matched with 102 pts in cohort 2. Demographic and baseline clinical characteristics were generally similar between the matched cohorts. A median OS from CGP report date of 16.5 months (95% CI 11.5-22.5) was observed for the WT-TRK pts, compared with 12.5 months (95% CI 9.5—not reached) for the pts with TRK fusion cancer. The proportional hazards ratio comparing pts with TRK fusion vs. WT-TRK cancer was 1.436 (95% CI 0.612-3.373). Conclusions: Pts with TRK fusion cancer had lower co-occurrence of other oncogenic biomarkers but had higher frequencies of MSI-high and TMB-high vs. pts with WT-TRK cancer. While no clear differences in OS were seen, there was a trend to shorter OS for pts with TRK fusion cancer. Further study allowing for inclusion of additional prognostic factors is needed to confirm the prognostic value of NTRK gene fusions. Citation Format: Lyudmila Bazhenova, Xiaolong Jiao, Andrew Lokker, Jeremy Snider, Emily Castellanos, Shivani Nanda, Virginia Fisher, Jihong Zong, Karen Keating, Marc Fellous. Cancers with NTRK gene fusions: Molecular characteristics and prognosis [abstract]. In: Proceedings of the AACR Special Conference on Advancing Precision Medicine Drug Development: Incorporation of Real-World Data and Other Novel Strategies; Jan 9-12, 2020; San Diego, CA. Philadelphia (PA): AACR; Clin Cancer Res 2020;26(12_Suppl_1):Abstract nr 09.