Sunyi Chi
Research Subjects
Sunyi Chi has a total of 12 co-authors. Here are the top co-authors by number of shared publications:
- Peng Wei
The University of Texas MD Anderson Cancer Center
3 shared publications - Xuelin Huang
Department of Biostatistics, The University of Texas MD Anderson Cancer Center, Houston, USA
2 shared publications - Bo Dong
College of Integrated Circuits and Optoelectronic Chips, Shenzhen Technology University, Shenzhen, China
1 shared publication - Christopher R. Flowers
UT MD Anderson Cancer Center, Houston, Texas, United States
1 shared publication - Chunlin Li
School of Statistics, University of Minnesota, Twin Cities
1 shared publication - Tianzhong Yang
Division of Biostatistics and Health Data Science, University of Minnesota, Twin Cities
1 shared publication
Latest Publications
- Open Access
MELODY: Mediation Analysis in Logistic Regression for High-Dimensional Mediators and a Binary Outcome
Mediation analysis is a pivotal tool for elucidating the indirect effect of an environmental factor or treatment on disease through potentially high-dimensional omics data, such as gene expression profiles. However, traditional mediation analysis methods tailored for binary outcomes often rely on the rare disease assumption in logistic regression and provide inadequate measures of total mediation effect when multiple mediators have effects in different directions. In this paper, we develop a MEdiation analysis framework in LOgistic regression for high-Dimensional mediators and a binarY outcome (MELODY). It leverages a second-moment-based measure analogous to the R 2 for linear models to quantify the total mediation effect. We also develop a variable selection procedure for high-dimensional data to reduce bias introduced by non-mediators. Our comprehensive simulations demonstrate the superior performance of MELODY in scenarios with non-rare disease binary outcomes and high-dimensional mediators. We apply MELODY to the Framingham Heart Study of over 5000 individuals to analyze the mediation effects of metabolomics and transcriptomics data on the pathways from sex to incident coronary heart disease.
Open AccessMELODY: Mediation Analysis in Logistic Regression for High-Dimensional Mediators and a Binary Outcome
Mediation analysis is a pivotal tool for elucidating the indirect effect of an environmental factor or treatment on disease through potentially high-dimensional omics data, such as gene expression profiles. However, traditional mediation analysis methods tailored for binary outcomes often rely on the rare disease assumption in logistic regression and provide inadequate measures of total mediation effect when multiple mediators have effects in different directions. In this paper, we develop a MEdiation analysis framework in LOgistic regression for high-Dimensional mediators and a binarY outcome (MELODY). It leverages a second-moment-based measure analogous to the R 2 for linear models to quantify the total mediation effect. We also develop a variable selection procedure for high-dimensional data to reduce bias introduced by non-mediators. Our comprehensive simulations demonstrate the superior performance of MELODY in scenarios with non-rare disease binary outcomes and high-dimensional mediators. We apply MELODY to the Framingham Heart Study of over 5000 individuals to analyze the mediation effects of metabolomics and transcriptomics data on the pathways from sex to incident coronary heart disease.
- Open Access
- 15 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1867-1874
APAM: Adaptive Pre-Training and Adaptive Meta Learning in Language Model for Noisy Labels and Long-Tailed Learning
Practical natural language processing (NLP) tasks often exhibit long-tailed distributions accompanied by noisy labels, posing significant challenges to the generalization and robustness of complex models such as Deep Neural Networks (DNNs). Traditional resampling techniques like oversampling or undersampling, while commonly employed, can easily lead to overfitting. A growing trend involves leveraging small amounts of metadata to learn data weights, alongside the demonstrated benefits of self-supervised pre-training, particularly for under-represented data. In this work, we propose a general framework that addresses both the long-tail and noisy label issues. Our model is adaptively tailored to the problem domain using a contrastive learning approach. The re-weighting module, a feed-forward network, learns explicit weighting functions and adjusts weights based on metadata. Additionally, our framework modifies the weights of terms in the loss function through a combination of polynomial expansion of the cross-entropy loss and focal loss. Extensive experiments consistently demonstrate the superior performance of our proposed framework compared to baseline methods. Finally, our comprehensive sensitivity analysis underscores the efficacy of the proposed framework in handling long-tailed problems and mitigating the adverse effects of noisy labels.
- 15 December 2023
- Published by Institute of Electrical and Electronics Engineers (IEEE)
- p. 1867-1874
Open AccessAPAM: Adaptive Pre-Training and Adaptive Meta Learning in Language Model for Noisy Labels and Long-Tailed Learning
Practical natural language processing (NLP) tasks often exhibit long-tailed distributions accompanied by noisy labels, posing significant challenges to the generalization and robustness of complex models such as Deep Neural Networks (DNNs). Traditional resampling techniques like oversampling or undersampling, while commonly employed, can easily lead to overfitting. A growing trend involves leveraging small amounts of metadata to learn data weights, alongside the demonstrated benefits of self-supervised pre-training, particularly for under-represented data. In this work, we propose a general framework that addresses both the long-tail and noisy label issues. Our model is adaptively tailored to the problem domain using a contrastive learning approach. The re-weighting module, a feed-forward network, learns explicit weighting functions and adjusts weights based on metadata. Additionally, our framework modifies the weights of terms in the loss function through a combination of polynomial expansion of the cross-entropy loss and focal loss. Extensive experiments consistently demonstrate the superior performance of our proposed framework compared to baseline methods. Finally, our comprehensive sensitivity analysis underscores the efficacy of the proposed framework in handling long-tailed problems and mitigating the adverse effects of noisy labels.
- Open Access
MASH: Mediation Analysis of Survival Outcome and High-dimensional Omics Mediators with Application to Complex Diseases
Environmental exposures such as cigarette smoking influence health out-comes through intermediate molecular phenotypes, such as the methylome, transcriptome, and metabolome. Mediation analysis is a useful tool for in-vestigating the role of potentially high-dimensional intermediate phenotypes in the relationship between environmental exposures and health outcomes. However, little work has been done on mediation analysis when the mediators are high-dimensional and the outcome is a survival endpoint, and none of it has provided a robust measure of total mediation effect. To this end, we propose an estimation procedure for M ediation A nalysis of S urvival outcome and H igh-dimensional omics mediators (MASH) based on sure independence screening for putative mediator variable selection and a second-moment-based measure of total mediation effect for survival data analogous to the R 2 measure in a linear model. Extensive simulations showed good performance of MASH in estimating the total mediation effect and identifying true mediators. By applying MASH to the metabolomics data of 1919 subjects in the Framingham Heart Study, we identified five metabolites as mediators of the effect of cigarette smoking on coronary heart disease risk (total mediation effect, 51.1%) and two metabolites as mediators between smoking and risk of cancer (total mediation effect, 50.7%). Application of MASH to a diffuse large B-cell lymphoma genomics data set identified copy-number variations for eight genes as mediators between the baseline International Prognostic Index score and overall survival.
Open AccessMASH: Mediation Analysis of Survival Outcome and High-dimensional Omics Mediators with Application to Complex Diseases
Environmental exposures such as cigarette smoking influence health out-comes through intermediate molecular phenotypes, such as the methylome, transcriptome, and metabolome. Mediation analysis is a useful tool for in-vestigating the role of potentially high-dimensional intermediate phenotypes in the relationship between environmental exposures and health outcomes. However, little work has been done on mediation analysis when the mediators are high-dimensional and the outcome is a survival endpoint, and none of it has provided a robust measure of total mediation effect. To this end, we propose an estimation procedure for M ediation A nalysis of S urvival outcome and H igh-dimensional omics mediators (MASH) based on sure independence screening for putative mediator variable selection and a second-moment-based measure of total mediation effect for survival data analogous to the R 2 measure in a linear model. Extensive simulations showed good performance of MASH in estimating the total mediation effect and identifying true mediators. By applying MASH to the metabolomics data of 1919 subjects in the Framingham Heart Study, we identified five metabolites as mediators of the effect of cigarette smoking on coronary heart disease risk (total mediation effect, 51.1%) and two metabolites as mediators between smoking and risk of cancer (total mediation effect, 50.7%). Application of MASH to a diffuse large B-cell lymphoma genomics data set identified copy-number variations for eight genes as mediators between the baseline International Prognostic Index score and overall survival.
- Open Access
Speeding up interval estimation for R 2 -based mediation effect of high-dimensional mediators via cross-fitting
Mediation analysis is a useful tool in investigating how molecular phenotypes such as gene expression mediate the effect of exposure on health outcomes. However, commonly used mean-based total mediation effect measures may suffer from cancellation of component-wise mediation effects in opposite directions in the presence of high-dimensional omics mediators. To overcome this limitation, we recently proposed a variance-based R-squared total mediation effect measure that relies on the computationally intensive nonparametric bootstrap for confidence interval estimation. In the work described herein, we formulated a more efficient two-stage, cross-fitted estimation procedure for the R 2 measure. To avoid potential bias, we performed iterative Sure Independence Screening (iSIS) in two subsamples to exclude the non-mediators, followed by ordinary least squares regressions for the variance estimation. We then constructed confidence intervals based on the newly derived closed-form asymptotic distribution of the R 2 measure. Extensive simulation studies demonstrated that this proposed procedure is much more computationally efficient than the resampling-based method, with comparable coverage probability. Furthermore, when applied to the Framingham Heart Study, the proposed method replicated the established finding of gene expression mediating age-related variation in systolic blood pressure and identified the role of gene expression profiles in the relationship between sex and high-density lipoprotein cholesterol level. The proposed estimation procedure is implemented in R package CFR2M .
Open AccessSpeeding up interval estimation for R 2 -based mediation effect of high-dimensional mediators via cross-fitting
Mediation analysis is a useful tool in investigating how molecular phenotypes such as gene expression mediate the effect of exposure on health outcomes. However, commonly used mean-based total mediation effect measures may suffer from cancellation of component-wise mediation effects in opposite directions in the presence of high-dimensional omics mediators. To overcome this limitation, we recently proposed a variance-based R-squared total mediation effect measure that relies on the computationally intensive nonparametric bootstrap for confidence interval estimation. In the work described herein, we formulated a more efficient two-stage, cross-fitted estimation procedure for the R 2 measure. To avoid potential bias, we performed iterative Sure Independence Screening (iSIS) in two subsamples to exclude the non-mediators, followed by ordinary least squares regressions for the variance estimation. We then constructed confidence intervals based on the newly derived closed-form asymptotic distribution of the R 2 measure. Extensive simulation studies demonstrated that this proposed procedure is much more computationally efficient than the resampling-based method, with comparable coverage probability. Furthermore, when applied to the Framingham Heart Study, the proposed method replicated the established finding of gene expression mediating age-related variation in systolic blood pressure and identified the role of gene expression profiles in the relationship between sex and high-density lipoprotein cholesterol level. The proposed estimation procedure is implemented in R package CFR2M .