A hybrid correlation analysis with application to imaging genetics

Wenxing Hu, Jian Fang, Vince Daniel Calhoun, Yu Ping Wang

Research output: Chapter in Book/Report/Conference proceedingConference contribution

Abstract

Investigating the association between brain regions and genes continues to be a challenging topic in imaging genetics. Current brain region of interest (ROI)-gene association studies normally reduce data dimension by averaging the value of voxels in each ROI. This averaging may lead to a loss of information due to the existence of functional sub-regions. Pearson correlation is widely used for association analysis. However, it only detects linear correlation whereas nonlinear correlation may exist among ROIs. In this work, we introduced distance correlation to ROI-gene association analysis, which can detect both linear and nonlinear correlations and overcome the limitation of averaging operations by taking advantage of the information at each voxel. Nevertheless, distance correlation usually has a much lower value than Pearson correlation. To address this problem, we proposed a hybrid correlation analysis approach, by applying canonical correlation analysis (CCA) to the distance covariance matrix instead of directly computing distance correlation. Incorporating CCA into distance correlation approach may be more suitable for complex disease study because it can detect highly associated pairs of ROI and gene groups, and may improve the distance correlation level and statistical power. In addition, we developed a novel nonlinear CCA, called distance kernel CCA, which seeks the optimal combination of features with the most significant dependence. This approach was applied to imaging genetic data from the Philadelphia Neurodevelopmental Cohort (PNC). Experiments showed that our hybrid approach produced more consistent results than conventional CCA across resampling and both the correlation and statistical significance were increased compared to distance correlation analysis. Further gene enrichment analysis and region of interest (ROI) analysis confirmed the associations of the identified genes with brain ROIs. Therefore, our approach provides a powerful tool for finding the correlation between brain imaging and genomic data.

Original languageEnglish (US)
Title of host publicationMedical Imaging 2018
Subtitle of host publicationImaging Informatics for Healthcare, Research, and Applications
PublisherSPIE
Volume10579
ISBN (Electronic)9781510616479
DOIs
StatePublished - Jan 1 2018
Externally publishedYes
EventMedical Imaging 2018: Imaging Informatics for Healthcare, Research, and Applications - Houston, United States
Duration: Feb 13 2018Feb 15 2018

Other

OtherMedical Imaging 2018: Imaging Informatics for Healthcare, Research, and Applications
CountryUnited States
CityHouston
Period2/13/182/15/18

    Fingerprint

Keywords

  • Canonical correlation
  • Distance correlation
  • Fmri
  • Imaging genetics
  • Kernel method

ASJC Scopus subject areas

  • Electronic, Optical and Magnetic Materials
  • Biomaterials
  • Atomic and Molecular Physics, and Optics
  • Radiology Nuclear Medicine and imaging

Cite this

Hu, W., Fang, J., Calhoun, V. D., & Wang, Y. P. (2018). A hybrid correlation analysis with application to imaging genetics. In Medical Imaging 2018: Imaging Informatics for Healthcare, Research, and Applications (Vol. 10579). [1057905] SPIE. https://doi.org/10.1117/12.2293556