Correlation analysis between MRI brain image and gene expression using ADNI data. Öffentlichkeit
Liu, Yajie (Spring 2022)
Published
Abstract
Background: With the development of medical imaging studies, numerous studies investigated the correlations with brain-related disease status, age, and the prognostic or diagnostic biomarkers for brain disease diagnosis [1]. There was a kind of regression called image-on-scalar model applied in this area. called image-on-scalar model aims to delineate the relationship between voxels or areas of interest (ROI) and a set of covariates of interest, such as demographic data, clinical features, and gene expression data, in which images were treated as a functional response variable [2].
Objectives: Examine the ability of elastic net regression in identifying genes that have an impact on brain MRI and examine whether there is any detectable correlation between age and MRI images.
Methods: Elastic net regression model was applied to examine the correlation. We built one model for each gene. In each model, we treated gene expression as dependent variable and image data as the independent variable, gender was tested as an independent variable. MRI image data, gene expression data, and age were from ADNI. And MRI image data were converted into 4232-dimension arrays by the auto-encoder method.
Results: We have selected the top 1000 genes with the highest variances to build elastic net regression models, and the top 25 models with the highest R-squares were analyzed. There were 12 of 25 models identified genes that were brain-related with the R-square of the model greater than 0.4. Models including gender had better performance than models that didn’t include gender. The R-square of examining the correlation between age and MRI image data had an R-square of about 0.45 and considered gender as a cofounder couldn’t improve the model performance.
Conclusions: In this study, we introduced an application of elastic net regression in identifying genes that have an impact on brain MRI, or in detecting correlation between age and MRI images. We found that the elastic net regression had relatively good in identifying genes that have an impact on brain MRI when considering gender effects. And the elastic net regression didn’t have a good performance in detecting the correlation between age and MRI image data.
Table of Contents
1. INTRODUCTION
2. METHODS
2.1 Data
2.1.1 Data source
2.1.2 Image data
2.1.3 Gene data
2.2 Statistical analysis
2.2.1 Elastic Net Regression
2.2.2 gender effect
3. RESULTS
3.1 Gene biomarker selection by elastic net
3.2 Gene biomarker selection by elastic net after removing gender effect
3.3 Age prediction by elastic net
4. DISCUSSION
REFERENCES
About this Master's Thesis
| School | |
|---|---|
| Department | |
| Subfield / Discipline | |
| Degree | |
| Submission | |
| Language |
|
| Research Field | |
| Stichwort | |
| Committee Chair / Thesis Advisor |
Primary PDF
| Thumbnail | Title | Date Uploaded | Actions |
|---|---|---|---|
|
|
Correlation analysis between MRI brain image and gene expression using ADNI data. () | 2022-04-11 12:53:23 -0400 |
|
Supplemental Files
| Thumbnail | Title | Date Uploaded | Actions |
|---|