Systematic Evaluation of Spatial Transcriptomics Alignment Methods Public

Abstract

Introduction: Accurate alignment of tissue sections is crucial for integrating spatial

transcriptomics with histological analyses. Current alignment methods often prioritize a single

modality—either gene expression or morphological features—which can be insufficient,

especially for distantly spaced sections. This study evaluates two alignment strategies, PASTE

and VALIS, to determine their effectiveness in aligning HER2+ breast cancer sections

characterized by heterogeneous spatial structures.

Methods: We applied PASTE and VALIS to align serial sections from HER2+ breast cancer

tissues. To assess alignment effectiveness, we developed a per-spot alignment cost function

incorporating bidirectional nearest-neighbor relationships and penalties for misalignments due to

reflection, scaling and rotation. This function quantitatively evaluates alignment performance

across both gene expression and morphological modalities.

Results: Our analysis revealed that PASTE frequently produced reflections, likely due to its gene

expression minimization process, leading to misalignments. Conversely, VALIS exhibited

scaling and rotation inconsistencies, particularly in regions where spatial landmarks were less

distinct. The per-spot alignment cost function effectively quantified these discrepancies,

highlighting the limitations of both methods in accurately aligning sections with weak spatial

landmarks.

Conclusion: The study highlights key limitations in current alignment strategies when applied to

tissue sections with high gene expression heterogeneity or lacking clear spatial landmarks. The

findings emphasize the necessity for advanced alignment approaches capable of integrating

multiple modalities and addressing spatial inconsistencies to enhance the accuracy and reliability

of spatial transcriptomics analyses.

Table of Contents

1. Introduction 1

2. Methods 5

2.1. Dataset used in this study 5

2.2. PASTE for Pairwise ST Section Alignment 6

2.3. VALIS for Pairwise WSI Registration 7

2.4. Evaluation metric: Per-Spot Cost 8

2.5. Validation of Alignment Costs via IRIS Joint Spatial Domain Partitions 10

2.6. Validation of Alignment Costs via Shared SVG Expression Patterns 11

2.7. Validation of Alignment Costs via Heatmap 12

2.8. Validation of Alignment Costs via Spot-level Cell-type Proportion 12

3. Results 14

3.1. Sequential Pairwise Alignment on High-Quality Cryosections 15

3.2. Pairwise Alignment of Cryosections with Increased Distance 18

3.3. Sequential Pairwise Alignment on Challenging Cryosections 21

4. Discussion 23

5. Bibliography 25

6. Appendix 27

6.1. Software and Package Utilization 27

6.2. Supplementary Figures 30

6.3. Supplementary Table 32

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	Rollins School of Public Health
Department	Biostatistics
Subfield / Discipline	Biostatistics - MPH & MSPH
Degree	M.S.P.H.
Submission	Master's Thesis
Language	English
Research Field	Biology, Biostatistics
Mot-clé	Spatial Transcriptomics Breast Cancer Study Spatial Domain Detection Genetics Biostatistics Unsupervised Machine Learning
Committee Chair / Thesis Advisor	Jian Hu, Emory University
Committee Members	Steve Qin, Emory University

Dernière modification

Primary PDF

Thumbnail	Title	Date Uploaded	Actions
	Systematic Evaluation of Spatial Transcriptomics Alignment Methods ()	2025-04-19 14:20:01 -0400	Press to Select an action Download

Supplemental Files

Thumbnail	Title	Date Uploaded	Actions