Single Cell Spatial Reconstruction Using Generative Models
![Thumbnail](http://memento.epfl.ch/image/28022/1440x810.jpg)
Event details
Date | 05.07.2024 |
Hour | 09:30 › 11:30 |
Speaker | Tingyang Yu |
Location | |
Category | Conferences - Seminars |
EDIC candidacy exam
Exam president: Prof. Pascal Frossard
Thesis advisor: Prof. Maria Brbic
Co-examiner: Prof. Gioele La Manno
Abstract
The field of molecular biology has been revolutionized by the advent of single-cell spatial transcriptomics (ST) techniques, which allow for the detailed analysis of transcriptomes while maintaining their spatial context within tissues. These techniques are essential for understanding the complex regulation of gene expression in relation to cellular microenvironments, providing crucial insights into biological processes and disease mechanisms. ST methods are generally divided into sequencing-based approaches, which capture the entire transcriptome but at a spot-level resolution, and imaging-based methods, which achieve single-cell resolution but are limited to predefined subsets of genes. This dichotomy underscores a significant limitation in current ST methodologies: the inability to capture a comprehensive transcriptome at single-cell spatial resolution simultaneously. To bridge this gap, recent research has focused on single-cell and spot alignment methods, as well as spatial coordinates reconstruction from single-cell RNA sequencing data. Notable methods include NovoSpaRc, Tangram, CytoSPACE, and CeLEry, each with its own limitations, such as reliance on predefined assumptions, data loss due to filtering, and scalability issues. In this work, we present to solve the problem using generative models which is designed to reconstruct single-cell spatial coordinates from gene expression profiles. It addresses the limitations of existing methods, providing a robust and versatile solution for spatial transcriptomics.
Background papers
Exam president: Prof. Pascal Frossard
Thesis advisor: Prof. Maria Brbic
Co-examiner: Prof. Gioele La Manno
Abstract
The field of molecular biology has been revolutionized by the advent of single-cell spatial transcriptomics (ST) techniques, which allow for the detailed analysis of transcriptomes while maintaining their spatial context within tissues. These techniques are essential for understanding the complex regulation of gene expression in relation to cellular microenvironments, providing crucial insights into biological processes and disease mechanisms. ST methods are generally divided into sequencing-based approaches, which capture the entire transcriptome but at a spot-level resolution, and imaging-based methods, which achieve single-cell resolution but are limited to predefined subsets of genes. This dichotomy underscores a significant limitation in current ST methodologies: the inability to capture a comprehensive transcriptome at single-cell spatial resolution simultaneously. To bridge this gap, recent research has focused on single-cell and spot alignment methods, as well as spatial coordinates reconstruction from single-cell RNA sequencing data. Notable methods include NovoSpaRc, Tangram, CytoSPACE, and CeLEry, each with its own limitations, such as reliance on predefined assumptions, data loss due to filtering, and scalability issues. In this work, we present to solve the problem using generative models which is designed to reconstruct single-cell spatial coordinates from gene expression profiles. It addresses the limitations of existing methods, providing a robust and versatile solution for spatial transcriptomics.
Background papers
Practical information
- General public
- Free