Efficient quantification of pattern similarity between spatial genomics and cell morphology Open Access

Yang, Yefeng (Spring 2024)

Permanent URL: https://etd.library.emory.edu/concern/etds/8336h340s?locale=en
Published

Abstract

Spatial transcriptomics (ST) is a powerful tool for studying gene expression patterns in spatial context, complemented with tissue histology image. Despite the availability of various methods to integrate the gene expression and tissue histology imaging in ST, the lack of a gold-standard metric to accurately quantify the similarity between features across different modalities has been a limitation for data integration. There is a strong demanding to measure the pattern similarity across features from different modalities, which provides insights into the association between molecular profile and cell morphology and enhances our understanding of tissue organization and function. To address this, we introduce DCOGC, a spatial pattern similarity measurement using a combination of correlation and mean squared error of orthogonal gradient curve that effectively captures the pattern similarity of features from different modality in the spatial context. Our method starts from identifying the common regions indicated by different modalities. Next, within each domain, it characterizes the pattern of each feature using gradient curves in two orthogonal directions. Then, we calculate the correlation and mean squared error of the different curves for each direction in every 

domain, summing them up to quantify the dissimilarity between two comparing features. The maximum value observed across all regions serves as an indicator of the highest level of feature disparity. DCOGC has been evaluated on multiple datasets, and compared with other popular spatial pattern measurements, i.e., SSIM, spatial correlation. The results demonstrate that our method can better identify image features that exhibit concordance with gene expression as well as those that show dissimilarity than all existing metrics. DCOGC can be further used to filter out image features dissimilarity to all gene expression patterns, which captured artificial effect such as blurriness resulting from loss of camera focus. DCOGC serves as a bridge between gene expression patterns and tissue histology imaging in ST, providing a robust metric to assess the similarity of features across diverse modalities

Table of Contents

optional

About this Master's Thesis

Rights statement
  • Permission granted by the author to include this thesis or dissertation in this repository. All rights reserved by the author. Please contact the author for information regarding the reproduction and use of this thesis or dissertation.
School
Department
Subfield / Discipline
Degree
Submission
Language
  • English
Research Field
Keyword
Committee Chair / Thesis Advisor
Committee Members
Last modified

Primary PDF

Supplemental Files