Gradient Boosted Trees for Spatial Data and Its Application to Medical Imaging Data.
Ontology highlight
ABSTRACT:
SUBMITTER: Iranzad R
PROVIDER: S-EPMC9615557 | biostudies-literature | 2022
REPOSITORIES: biostudies-literature
Publications
Gradient Boosted Trees for Spatial Data and Its Application to Medical Imaging Data.
Iranzad Reza R Liu Xiao X Chaovalitwongse W Art WA Hippe Daniel D Wang Shouyi S Han Jie J Thammasorn Phawis P Duan Chunyan C Zeng Jing J Bowen Stephen S
IISE transactions on healthcare systems engineering 20211109 3
Boosting Trees are one of the most successful statistical learning approaches that involve sequentially growing an ensemble of simple regression trees ("weak learners"). This paper proposes a gradient Boosted Trees algorithm for Spatial Data (Boost-S) with covariate information. Boost-S integrates the spatial correlation into the classical framework of eXtreme Gradient Boosting. Each tree is constructed by solving a regularized optimization problem, where the objective function takes into accoun ...[more]
Similar Datasets
Project description:BackgroundWaist circumference is becoming recognized as a useful predictor of health risks in clinical research. However, clinical datasets tend to lack this measurement and self-reported values tend to be inaccurate. Predicting waist circumference from standard physical features could be a viable method for generating this information when it is missing or mitigating the impact of inaccurate self-reports. This study determined the degree to which the XGBoost advanced machine learning algorithm could build models that predict waist circumference from height, weight, calculated Body Mass Index, age, race/ethnicity and sex, whether they perform better than current models based on linear regression, and the relative importance of each feature in this prediction.MethodsWe trained tree-based models (via XGBoost gradient boosting) and linear models (via regression) to predict waist circumference from height, weight, Body Mass Index, age, race/ethnicity and sex (n = 60,740 participants). We created 10 iterations of each model, each using 90% of the dataset for training and the remaining 10% for testing performance (this group was different for each iteration). We calculated model performance and feature importance as an average across 10 iterations. We then externally validated the ensembled version of the top model.ResultsThe XGBoost model predicted waist circumference with a mean bias ± standard deviation of 0.0 ± 0.04 cm and a root mean squared error of 4.7 ± 0.05 cm, with performance varying slightly by sex and race/ethnicity. The XGBoost model showed varying degrees of improvement over linear regression models. The top 3 predictors were Body Mass Index, weight and race (Asian). External validation found that on average this model overestimated waist circumference by 4.65 cm in the United Kingdom population (mainly due to overprediction in females) and underestimated waist circumference by 1.7 cm in the Chinese population. The respective root mean squared errors were 7.7 cm and 7.1 cm.ConclusionsXGBoost-based models accurately predict waist circumference from standard physical features. Waist circumference prediction using this approach would be valuable for epidemiological research and beyond.
| S-EPMC7944598 | biostudies-literature
Project description:Protein-RNA interactions play essential roles in many biological aspects. Quantifying the binding affinity of protein-RNA complexes is helpful to the understanding of protein-RNA recognition mechanisms and identification of strong binding partners. Due to experimentally measured protein-RNA binding affinity data available is still limited to date, there is a pressing demand for accurate and reliable computational approaches. In this paper, we propose a computational approach, PredPRBA, which can effectively predict protein-RNA binding affinity using gradient boosted regression trees. We build a dataset of protein-RNA binding affinity that includes 103 protein-RNA complex structures manually collected from related literature. Then, we generate 37 kinds of sequence and structural features and explore the relationship between the features and protein-RNA binding affinity. We find that the binding affinity mainly depends on the structure of RNA molecules. According to the type of RNA associated with proteins composed of the protein-RNA complex, we split the 103 protein-RNA complexes into six categories. For each category, we build a gradient boosted regression tree (GBRT) model based on the generated features. We perform a comprehensive evaluation for the proposed method on the binding affinity dataset using leave-one-out cross-validation. We show that PredPRBA achieves correlations ranging from 0.723 to 0.897 among six categories, which is significantly better than other typical regression methods and the pioneer protein-RNA binding affinity predictor SPOT-Seq-RNA. In addition, a user-friendly web server has been developed to predict the binding affinity of protein-RNA complexes. The PredPRBA webserver is freely available at http://PredPRBA.denglab.org/.
| S-EPMC6688581 | biostudies-literature
Project description:Pseudomonas syringae is a genetically diverse bacterial species complex responsible for numerous agronomically important crop diseases. Individual P. syringae isolates are assigned pathovar designations based on their host of isolation and the associated disease symptoms, and these pathovar designations are often assumed to reflect host specificity although this assumption has rarely been rigorously tested. Here we developed a rapid seed infection assay to measure the virulence of 121 diverse P. syringae isolates on common bean (Phaseolus vulgaris). This collection includes P. syringae phylogroup 2 (PG2) bean isolates (pathovar syringae) that cause bacterial spot disease and P. syringae phylogroup 3 (PG3) bean isolates (pathovar phaseolicola) that cause the more serious halo blight disease. We found that bean isolates in general were significantly more virulent on bean than non-bean isolates and observed no significant virulence difference between the PG2 and PG3 bean isolates. However, when we compared virulence within PGs we found that PG3 bean isolates were significantly more virulent than PG3 non-bean isolates, while there was no significant difference in virulence between PG2 bean and non-bean isolates. These results indicate that PG3 strains have a higher level of host specificity than PG2 strains. We then used gradient boosting machine learning to predict each strain's virulence on bean based on whole genome k-mers, type III secreted effector k-mers, and the presence/absence of type III effectors and phytotoxins. Our model performed best using whole genome data and was able to predict virulence with high accuracy (mean absolute error = 0.05). Finally, we functionally validated the model by predicting virulence for 16 strains and found that 15 (94%) had virulence levels within the bounds of estimated predictions. This study strengthens the hypothesis that P. syringae PG2 strains have evolved a different lifestyle than other P. syringae strains as reflected in their lower level of host specificity. It also acts as a proof-of-principle to demonstrate the power of machine learning for predicting host specific adaptation.
| S-EPMC9352200 | biostudies-literature
Project description:PurposeHyperspectral imaging has the potential to improve intraoperative decision making if tissue characterisation is performed in real-time and with high-resolution. Hyperspectral snapshot mosaic sensors offer a promising approach due to their fast acquisition speed and compact size. However, a demosaicking algorithm is required to fully recover the spatial and spectral information of the snapshot images. Most state-of-the-art demosaicking algorithms require ground-truth training data with paired snapshot and high-resolution hyperspectral images, but such imagery pairs with the exact same scene are physically impossible to acquire in intraoperative settings. In this work, we present a fully unsupervised hyperspectral image demosaicking algorithm which only requires exemplar snapshot images for training purposes.MethodsWe regard hyperspectral demosaicking as an ill-posed linear inverse problem which we solve using a deep neural network. We take advantage of the spectral correlation occurring in natural scenes to design a novel inter spectral band regularisation term based on spatial gradient consistency. By combining our proposed term with standard regularisation techniques and exploiting a standard data fidelity term, we obtain an unsupervised loss function for training deep neural networks, which allows us to achieve real-time hyperspectral image demosaicking.ResultsQuantitative results on hyperspetral image datasets show that our unsupervised demosaicking approach can achieve similar performance to its supervised counter-part, and significantly outperform linear demosaicking. A qualitative user study on real snapshot hyperspectral surgical images confirms the results from the quantitative analysis.ConclusionOur results suggest that the proposed unsupervised algorithm can achieve promising hyperspectral demosaicking in real-time thus advancing the suitability of the modality for intraoperative use.
| S-EPMC10284955 | biostudies-literature
Project description:Cellular anatomy and signaling vary across niches, which can induce gradated gene expressions in subpopulations of cells. Such spatial transcriptomic gradient (STG) makes a significant source of intratumor heterogeneity and can influence tumor invasion, progression, and response to treatment. Here we report Local Spatial Gradient Inference (LSGI), a computational framework that systematically identifies spatial locations with prominent, interpretable STGs from spatial transcriptomic (ST) data. To achieve so, LSGI scrutinizes each sliding window employing non-negative matrix factorization (NMF) combined with linear regression. With LSGI, we demonstrated the identification of spatially proximal yet opposite directed pathway gradients in a glioblastoma dataset. We further applied LSGI to 87 tumor ST datasets reported from nine published studies and identified both pan-cancer and tumor-type specific pathways with gradated expression patterns, such as epithelial mesenchymal transition, MHC complex, and hypoxia. The local gradients were further categorized according to their association to tumor-TME (tumor microenvironment) interface, highlighting the pathways related to spatial transcriptional intratumoral heterogeneity. We conclude that LSGI enables highly interpretable STG analysis which can reveal novel insights in tumor biology from the increasingly reported tumor ST datasets.
| S-EPMC10983986 | biostudies-literature
Project description:BackgroundClinical models to predict first trimester viability are traditionally based on multivariable logistic regression (LR) which is not directly interpretable for non-statistical experts like physicians. Furthermore, LR requires complete datasets and pre-established variables specifications. In this study, we leveraged the internal non-linearity, feature selection and missing values handling mechanisms of machine learning algorithms, along with a post-hoc interpretability strategy, as potential advantages over LR for clinical modeling.MethodsThe dataset included 1154 patients with 2377 individual scans and was obtained from a prospective observational cohort study conducted at a hospital in London, UK, from March 2014 to May 2019. The data were split into a training (70%) and a test set (30%). Parsimonious and complete multivariable models were developed from two algorithms to predict first trimester viability at 11-14 weeks gestational age (GA): LR and light gradient boosted machine (LGBM). Missing values were handled by multiple imputation where appropriate. The SHapley Additive exPlanations (SHAP) framework was applied to derive individual explanations of the models.ResultsThe parsimonious LGBM model had similar discriminative and calibration performance as the parsimonious LR (AUC 0.885 vs 0.860; calibration slope: 1.19 vs 1.18). The complete models did not outperform the parsimonious models. LGBM was robust to the presence of missing values and did not require multiple imputation unlike LR. Decision path plots and feature importance analysis revealed different algorithm behaviors despite similar predictive performance. The main driving variable from the LR model was the pre-specified interaction between fetal heart presence and mean sac diameter. The crown-rump length variable and a proxy variable reflecting the difference in GA between expected and observed GA were the two most important variables of LGBM. Finally, while variable interactions must be specified upfront with LR, several interactions were ranked by the SHAP framework among the most important features learned automatically by the LGBM algorithm.ConclusionsGradient boosted algorithms performed similarly to carefully crafted LR models in terms of discrimination and calibration for first trimester viability prediction. By handling multi-collinearity, missing values, feature selection and variable interactions internally, the gradient boosted trees algorithm, combined with SHAP, offers a serious alternative to traditional LR models.
| S-EPMC8674730 | biostudies-literature
Project description:Substances that do not degrade over time have proven to be harmful to the environment and are dangerous to living organisms. Being able to predict the biodegradability of substances without costly experiments is useful. Recently, the quantitative structure-activity relationship (QSAR) models have proposed effective solutions to this problem. However, the molecular descriptor datasets usually suffer from the problems of unbalanced class distribution, which adversely affects the efficiency and generalization of the derived models. Accordingly, this study aims at validating the performances of balanced random trees (RTs) and boosted C5.0 decision trees (DTs) to construct QSAR models to classify the ready biodegradation of substances and their abilities to deal with unbalanced data. The balanced RTs model algorithm builds individual trees using balanced bootstrap samples, while the boosted C5.0 DT is modeled using cost-sensitive learning. We employed the two-dimensional molecular descriptor dataset, which is publicly available through the University of California, Irvine (UCI) machine learning repository. The molecular descriptors were ranked according to their contributions to the balanced RTs classification process. The performance of the proposed models was compared with previously reported results. Based on the statistical measures, the experimental results showed that the proposed models outperform the classification results of the support vector machine (SVM), K-nearest neighbors (KNN), and discrimination analysis (DA). Classification measures were analyzed in terms of accuracy, sensitivity, specificity, precision, false positive rate, false negative rate, F1 score, receiver operating characteristic (ROC) curve, and area under the ROC curve (AUROC).
| S-EPMC7763457 | biostudies-literature
Project description:In spatial transcriptomics (ST) data, biologically relevant features such as tissue compartments or cell-state transitions are reflected by gene expression gradients. Here, we present SpaceWalker, a visual analytics tool for exploring the local gradient structure of 2D and 3D ST data. The user can be guided by the local intrinsic dimensionality of the high-dimensional data to define seed locations, from which a flood-fill algorithm identifies transcriptomically similar cells on the fly, based on the high-dimensional data topology. In several use cases, we demonstrate that the spatial projection of these flooded cells highlights tissue architectural features and that interactive retrieval of gene expression gradients in the spatial and transcriptomic domains confirms known biology. We also show that SpaceWalker generalizes to several different ST protocols and scales well to large, multi-slice, 3D whole-brain ST data while maintaining real-time interaction performance.
| S-EPMC10753200 | biostudies-literature
Project description:BackgroundThe 2013-2016 West African Ebola epidemic has been the largest to date with >11 000 deaths in the affected countries. The data collected have provided more insight into the case fatality ratio (CFR) and how it varies with age and other characteristics. However, the accuracy and precision of the naive CFR remain limited because 44% of survival outcomes were unreported.MethodsUsing a boosted regression tree model, we imputed survival outcomes (ie, survival or death) when unreported, corrected for model imperfection to estimate the CFR without imputation, with imputation, and adjusted with imputation. The method allowed us to further identify and explore relevant clinical and demographic predictors of the CFR.ResultsThe out-of-sample performance (95% confidence interval [CI]) of our model was good: sensitivity, 69.7% (52.5-75.6%); specificity, 69.8% (54.1-75.6%); percentage correctly classified, 69.9% (53.7-75.5%); and area under the receiver operating characteristic curve, 76.0% (56.8-82.1%). The adjusted CFR estimates (95% CI) for the 2013-2016 West African epidemic were 82.8% (45.6-85.6%) overall and 89.1% (40.8-91.6%), 65.6% (61.3-69.6%), and 79.2% (45.4-84.1%) for Sierra Leone, Guinea, and Liberia, respectively. We found that district, hospitalisation status, age, case classification, and quarter (date of case reporting aggregated at three-month intervals) explained 93.6% of the variance in the naive CFR.ConclusionsThe adjusted CFR estimates improved the naive CFR estimates obtained without imputation and were more representative. Used in conjunction with other resources, adjusted estimates will inform public health contingency planning for future Ebola epidemics, and help better allocate resources and evaluate the effectiveness of future inventions.
| S-EPMC7286386 | biostudies-literature
Project description:In savannas, predicting how vegetation varies is a longstanding challenge. Spatial patterning in vegetation may structure that variability, mediated by spatial interactions, including competition and facilitation. Here, we use unique high-resolution, spatially extensive data of tree distributions in an African savanna, derived from airborne Light Detection and Ranging (LiDAR), to examine tree-clustering patterns. We show that tree cluster sizes were governed by power laws over two to three orders of magnitude in spatial scale and that the parameters on their distributions were invariant with respect to underlying environment. Concluding that some universal process governs spatial patterns in tree distributions may be premature. However, we can say that, although the tree layer may look unpredictable locally, at scales relevant to prediction in, e.g., global vegetation models, vegetation is instead strongly structured by regular statistical distributions.
| S-EPMC6561214 | biostudies-literature