A Multilevel Bayesian Approach to Improve Effect Size Estimation in Regression Modeling of Metabolomics Data Utilizing Imputation with Uncertainty
To ensure scientific reproducibility of metabolomics data, alternative statistical methods are needed. A paradigm shift away from the <i>p</i>-value toward an embracement of uncertainty and interval estimation of a metabolite’s true effect size may lead to improved study design and great...
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
MDPI AG
2020-08-01
|
| Series: | Metabolites |
| Subjects: | |
| Online Access: | https://www.mdpi.com/2218-1989/10/8/319 |
| _version_ | 1797559978706337792 |
|---|---|
| author | Christopher E. Gillies Theodore S. Jennaro Michael A. Puskarich Ruchi Sharma Kevin R. Ward Xudong Fan Alan E. Jones Kathleen A. Stringer |
| author_facet | Christopher E. Gillies Theodore S. Jennaro Michael A. Puskarich Ruchi Sharma Kevin R. Ward Xudong Fan Alan E. Jones Kathleen A. Stringer |
| author_sort | Christopher E. Gillies |
| collection | DOAJ |
| description | To ensure scientific reproducibility of metabolomics data, alternative statistical methods are needed. A paradigm shift away from the <i>p</i>-value toward an embracement of uncertainty and interval estimation of a metabolite’s true effect size may lead to improved study design and greater reproducibility. Multilevel Bayesian models are one approach that offer the added opportunity of incorporating imputed value uncertainty when missing data are present. We designed simulations of metabolomics data to compare multilevel Bayesian models to standard logistic regression with corrections for multiple hypothesis testing. Our simulations altered the sample size and the fraction of significant metabolites truly different between two outcome groups. We then introduced missingness to further assess model performance. Across simulations, the multilevel Bayesian approach more accurately estimated the effect size of metabolites that were significantly different between groups. Bayesian models also had greater power and mitigated the false discovery rate. In the presence of increased missing data, Bayesian models were able to accurately impute the <i>true</i> concentration and incorporating the uncertainty of these estimates improved overall prediction. In summary, our simulations demonstrate that a multilevel Bayesian approach accurately quantifies the estimated effect size of metabolite predictors in regression modeling, particularly in the presence of missing data. |
| first_indexed | 2024-03-10T17:52:55Z |
| format | Article |
| id | doaj.art-7bd5a4ac2a32402ea6913bafb9d4eb74 |
| institution | Directory Open Access Journal |
| issn | 2218-1989 |
| language | English |
| last_indexed | 2024-03-10T17:52:55Z |
| publishDate | 2020-08-01 |
| publisher | MDPI AG |
| record_format | Article |
| series | Metabolites |
| spelling | doaj.art-7bd5a4ac2a32402ea6913bafb9d4eb742023-11-20T09:17:07ZengMDPI AGMetabolites2218-19892020-08-0110831910.3390/metabo10080319A Multilevel Bayesian Approach to Improve Effect Size Estimation in Regression Modeling of Metabolomics Data Utilizing Imputation with UncertaintyChristopher E. Gillies0Theodore S. Jennaro1Michael A. Puskarich2Ruchi Sharma3Kevin R. Ward4Xudong Fan5Alan E. Jones6Kathleen A. Stringer7Department of Emergency Medicine, University of Michigan, Ann Arbor, MI 48109, USADepartment of Clinical Pharmacy, College of Pharmacy, University of Michigan, Ann Arbor, MI 48109, USADepartment of Emergency Medicine, University of Minnesota, Minneapolis, MN 55455, USADepartment of Biomedical Engineering, University of Michigan, Ann Arbor, MI 48109, USADepartment of Emergency Medicine, University of Michigan, Ann Arbor, MI 48109, USAMichigan Center for Integrative Research in Critical Care (MCIRCC), University of Michigan, Ann Arbor, MI 48109, USADepartment of Emergency Medicine, University of Mississippi Medical Center, Jackson, MS 39216, USAMichigan Center for Integrative Research in Critical Care (MCIRCC), University of Michigan, Ann Arbor, MI 48109, USATo ensure scientific reproducibility of metabolomics data, alternative statistical methods are needed. A paradigm shift away from the <i>p</i>-value toward an embracement of uncertainty and interval estimation of a metabolite’s true effect size may lead to improved study design and greater reproducibility. Multilevel Bayesian models are one approach that offer the added opportunity of incorporating imputed value uncertainty when missing data are present. We designed simulations of metabolomics data to compare multilevel Bayesian models to standard logistic regression with corrections for multiple hypothesis testing. Our simulations altered the sample size and the fraction of significant metabolites truly different between two outcome groups. We then introduced missingness to further assess model performance. Across simulations, the multilevel Bayesian approach more accurately estimated the effect size of metabolites that were significantly different between groups. Bayesian models also had greater power and mitigated the false discovery rate. In the presence of increased missing data, Bayesian models were able to accurately impute the <i>true</i> concentration and incorporating the uncertainty of these estimates improved overall prediction. In summary, our simulations demonstrate that a multilevel Bayesian approach accurately quantifies the estimated effect size of metabolite predictors in regression modeling, particularly in the presence of missing data.https://www.mdpi.com/2218-1989/10/8/319hierarchical modelingnuclear magnetic resonance spectroscopyBayesian statisticsmissing valuesimputationmultiple test corrections |
| spellingShingle | Christopher E. Gillies Theodore S. Jennaro Michael A. Puskarich Ruchi Sharma Kevin R. Ward Xudong Fan Alan E. Jones Kathleen A. Stringer A Multilevel Bayesian Approach to Improve Effect Size Estimation in Regression Modeling of Metabolomics Data Utilizing Imputation with Uncertainty Metabolites hierarchical modeling nuclear magnetic resonance spectroscopy Bayesian statistics missing values imputation multiple test corrections |
| title | A Multilevel Bayesian Approach to Improve Effect Size Estimation in Regression Modeling of Metabolomics Data Utilizing Imputation with Uncertainty |
| title_full | A Multilevel Bayesian Approach to Improve Effect Size Estimation in Regression Modeling of Metabolomics Data Utilizing Imputation with Uncertainty |
| title_fullStr | A Multilevel Bayesian Approach to Improve Effect Size Estimation in Regression Modeling of Metabolomics Data Utilizing Imputation with Uncertainty |
| title_full_unstemmed | A Multilevel Bayesian Approach to Improve Effect Size Estimation in Regression Modeling of Metabolomics Data Utilizing Imputation with Uncertainty |
| title_short | A Multilevel Bayesian Approach to Improve Effect Size Estimation in Regression Modeling of Metabolomics Data Utilizing Imputation with Uncertainty |
| title_sort | multilevel bayesian approach to improve effect size estimation in regression modeling of metabolomics data utilizing imputation with uncertainty |
| topic | hierarchical modeling nuclear magnetic resonance spectroscopy Bayesian statistics missing values imputation multiple test corrections |
| url | https://www.mdpi.com/2218-1989/10/8/319 |
| work_keys_str_mv | AT christopheregillies amultilevelbayesianapproachtoimproveeffectsizeestimationinregressionmodelingofmetabolomicsdatautilizingimputationwithuncertainty AT theodoresjennaro amultilevelbayesianapproachtoimproveeffectsizeestimationinregressionmodelingofmetabolomicsdatautilizingimputationwithuncertainty AT michaelapuskarich amultilevelbayesianapproachtoimproveeffectsizeestimationinregressionmodelingofmetabolomicsdatautilizingimputationwithuncertainty AT ruchisharma amultilevelbayesianapproachtoimproveeffectsizeestimationinregressionmodelingofmetabolomicsdatautilizingimputationwithuncertainty AT kevinrward amultilevelbayesianapproachtoimproveeffectsizeestimationinregressionmodelingofmetabolomicsdatautilizingimputationwithuncertainty AT xudongfan amultilevelbayesianapproachtoimproveeffectsizeestimationinregressionmodelingofmetabolomicsdatautilizingimputationwithuncertainty AT alanejones amultilevelbayesianapproachtoimproveeffectsizeestimationinregressionmodelingofmetabolomicsdatautilizingimputationwithuncertainty AT kathleenastringer amultilevelbayesianapproachtoimproveeffectsizeestimationinregressionmodelingofmetabolomicsdatautilizingimputationwithuncertainty AT christopheregillies multilevelbayesianapproachtoimproveeffectsizeestimationinregressionmodelingofmetabolomicsdatautilizingimputationwithuncertainty AT theodoresjennaro multilevelbayesianapproachtoimproveeffectsizeestimationinregressionmodelingofmetabolomicsdatautilizingimputationwithuncertainty AT michaelapuskarich multilevelbayesianapproachtoimproveeffectsizeestimationinregressionmodelingofmetabolomicsdatautilizingimputationwithuncertainty AT ruchisharma multilevelbayesianapproachtoimproveeffectsizeestimationinregressionmodelingofmetabolomicsdatautilizingimputationwithuncertainty AT kevinrward multilevelbayesianapproachtoimproveeffectsizeestimationinregressionmodelingofmetabolomicsdatautilizingimputationwithuncertainty AT xudongfan multilevelbayesianapproachtoimproveeffectsizeestimationinregressionmodelingofmetabolomicsdatautilizingimputationwithuncertainty AT alanejones multilevelbayesianapproachtoimproveeffectsizeestimationinregressionmodelingofmetabolomicsdatautilizingimputationwithuncertainty AT kathleenastringer multilevelbayesianapproachtoimproveeffectsizeestimationinregressionmodelingofmetabolomicsdatautilizingimputationwithuncertainty |