Juan Pablo Lewinger

Juan Pablo Lewinger

Auto-generated profile, based on publicly available information

Juan Pablo Lewinger

University of Southern California 2 Division of Biostatistics, Department of Population and Public Health Sciences Los Angeles, California

Research Focus & Metrics

Academic Career Length

22Years (2005-Present)

(Calculated once Scilit publication data is synced: current year − year of first publication + 1)

Representative Publications

View All Publications
24 July 2007 in Genetic Epidemiology
Verified
Journal Article
Closed access
PDF file
JLJuan Pablo Lewinger
Juan Pablo Lewinger
DCDavid V. Conti
David V. Conti
JBJames W. Baurley
James W. Baurley
TTTimothy J. Triche
Timothy J. Triche
DTDuncan C. Thomas
Duncan C. Thomas

17 February 2006 in Statistics in Medicine
Verified
Journal Article
Closed access
PDF file
SBShelley B. Bull
Shelley B. Bull
JLJuan Pablo Lewinger
Juan Pablo Lewinger
SLSophia S. F. Lee
Sophia S. F. Lee

10 October 2018 in BMC Bioinformatics
Verified
Journal Article
Open access
PDF file
JLJie Liu
Jie Liu
GLGangning Liang
Gangning Liang
KSKimberly D Siegmund
Kimberly D Siegmund
JLJuan Pablo Lewinger
Juan Pablo Lewinger
To integrate molecular features from multiple high-throughput platforms in prediction, a regression model that penalizes features from all platforms equally is commonly used. However, data from different platforms are likely to differ in effect sizes, the proportion of predictive features, and correlations structures. Subtle but important features may be missed by shrinking all features equally. We propose an Elastic net (EN) model with separate tuning parameter penalties for each platform that is fit using standard software. In a comprehensive simulation study, we evaluated the performance of EN logistic regression with multiple tuning penalties. We found that when the number of informative features differs among the platforms, and when there is no notable correlation between the features from different platforms, the multi-tuning parameter EN yields more predictive models. Moreover, the multi-tuning parameter EN is robust, in the sense that there is no loss of predictivity relative to a single tuning parameter EN when features across all platforms have similar effects. We also investigated the performance of multi-tuning parameter EN using real cancer datasets. The proposed multi-tuning parameter EN model, fit using standard penalized regression software, can achieve better prediction in sample classification when integrating multiple genomic platforms, compared to the traditional method where a single penalty parameter is used for all features in different platforms.

1 January 2013 in Frontiers in Endocrinology
Verified
Journal Article
Open access
PDF file
KBKimberly Banks
Kimberly Banks
ETEllenie Tuazon
Ellenie Tuazon
KBKiros Berhane
Kiros Berhane
CKChester J. Koh
Chester J. Koh
RFRoger E. De Filippo
Roger E. De Filippo
VCVictoria K. Cortessis
Victoria K. Cortessis
Introduction: Risk of testicular germ cell tumors (TGCT) is consistently associated with a history of cryptorchidism (CO) in epidemiologic studies. Factors modifying the association may provide insights regarding etiology of TGCT and suggest a basis for individualized care of CO. To identify modifiers of the CO-TGCT association, we conducted a comprehensive, quantitative evaluation of epidemiologic data. Materials and Methods: Human studies cited in PubMed or ISI Web of Science indices through December 2011 and selected unpublished epidemiologic data were reviewed to identify 35 articles and one unpublished dataset with high-quality data on the CO-TGCT association. Association data were extracted as point and 95% confidence interval estimates of odds ratio (OR) or standardized incidence ratio (SIR), or as tabulated data. Values were recorded for each study population, and for subgroups defined by features of study design, CO and TGCT. Extracted data were used to estimate summary risk ratios (sRR) and evaluate heterogeneity of the CO-TGCT association between subgroups. Results: The overall meta-analysis showed that history of CO is associated with four-fold increased TGCT risk [RR = 4.1(95% CI = 3.6–4.7)]. Subgroup analyses identified five determinants of stronger association: bilateral CO, unilateral CO ipsilateral to TGCT, delayed CO treatment, TGCT diagnosed before 1970, and seminoma histology. Conclusions: Modifying factors may provide insight into TGCT etiology and suggest improved approaches to managing CO. Based on available data, CO patients and their parents or caregivers should be made aware of elevated TGCT risk following orchidopexy, regardless of age at repair, unilateral vs. bilateral non-descent, or position of undescended testes.

30 April 2013 in Genetic Epidemiology
Verified
Journal Article
Closed access
PDF file
JLJuan Pablo Lewinger
Juan Pablo Lewinger
JMJohn L. Morrison
John L. Morrison
DTDuncan C. Thomas
Duncan C. Thomas
CMCassandra E. Murcray
Cassandra E. Murcray
DCDavid V. Conti
David V. Conti
WGW. James Gauderman
W. James Gauderman

11 September 2020 in Bioinformatics
Verified
Journal Article
Open access
PDF file
CZChubing Zeng
Chubing Zeng
DTDuncan Campbell Thomas
Duncan Campbell Thomas
JLJuan Pablo Lewinger
Juan Pablo Lewinger
Associated with genomic features like gene expression, methylation and genotypes, used in statistical modeling of health outcomes, there is a rich set of meta-features like functional annotations, pathway information and knowledge from previous studies, that can be used post hoc to facilitate the interpretation of a model. However, using this meta-feature information a priori rather than post hoc can yield improved prediction performance as well as enhanced model interpretation. We propose a new penalized regression approach that allows a priori integration of external meta-features. The method extends LASSO regression by incorporating individualized penalty parameters for each regression coefficient. The penalty parameters are, in turn, modeled as a log-linear function of the meta-features and are estimated from the data using an approximate empirical Bayes approach. Optimization of the marginal likelihood on which the empirical Bayes estimation is performed using a fast and stable majorization–minimization procedure. Through simulations, we show that the proposed regression with individualized penalties can outperform the standard LASSO in terms of both parameters estimation and prediction performance when the external data is informative. We further demonstrate our approach with applications to gene expression studies of bone density and breast cancer. The methods have been implemented in the R package xtune freely available for download from https://cran.r-project.org/web/packages/xtune/index.html.