Multistage Optimal Transport evaluated on scRNA-seq Human Bone Marrow data
(2026)Department of Automatic Control
- Abstract
- With novel techniques like single-cell RNA sequencing (scRNA-seq) we can look inside single cells in great detail and measure their individual gene activity by counting the mRNA molecules of the cells. In order to model cell differentiation one would naturally turn to longitudinal data of single cells. However, snapshot datasets is the basis of single cell studies due to the way the datasets are collected. The advantage with scRNA-seq is that samples are very informative and in healthy bone marrow, often contain a good representation of cells in different stages of the differentiation process. We can therefore assume that we have sufficient information to create a model for the differentiation process using the snapshot data alongside... (More)
- With novel techniques like single-cell RNA sequencing (scRNA-seq) we can look inside single cells in great detail and measure their individual gene activity by counting the mRNA molecules of the cells. In order to model cell differentiation one would naturally turn to longitudinal data of single cells. However, snapshot datasets is the basis of single cell studies due to the way the datasets are collected. The advantage with scRNA-seq is that samples are very informative and in healthy bone marrow, often contain a good representation of cells in different stages of the differentiation process. We can therefore assume that we have sufficient information to create a model for the differentiation process using the snapshot data alongside optimal transport frameworks. Building on the assumption that cell states similar to each other have similar gene expression, and that this gene expression can provide information about the order of these cells in a differentiation process, we use a newly proposed optimal transport algorithm to compute probable transitions between cell states as well as a pseudotime ordering of the cells. We use these transitions to extract subsets of cells which represent differentiation trajectories through human bone marrow samples, and order the subset cells in pseudotime. We then estimate expressions of some common marker genes across the subset’s pseudotime in order to verify that the model captures expected biological behaviour. We also apply the method to a sample of cells connected to Acute Myeloid Leukemia (AML) to observe if the changed biological content causes these marker genes to be expressed differently in the sample. The ambition is that this method can aid in further study of the development of AML. (Less)
Please use this url to cite or link to this publication:
https://lup.lub.lu.se/student-papers/record/9249051
- author
- Conradi, Vilhelm and Nelsson, Hugo
- supervisor
-
- Karl Rydén LU
- Anders Rantzer LU
- Magnus Fontes LU
- organization
- year
- 2026
- type
- H3 - Professional qualifications (4 Years - )
- subject
- report number
- TFRT-6318
- other publication id
- 0280-5316
- language
- English
- id
- 9249051
- date added to LUP
- 2026-08-25 15:17:49
- date last changed
- 2026-08-25 15:17:49
@misc{9249051,
abstract = {{With novel techniques like single-cell RNA sequencing (scRNA-seq) we can look inside single cells in great detail and measure their individual gene activity by counting the mRNA molecules of the cells. In order to model cell differentiation one would naturally turn to longitudinal data of single cells. However, snapshot datasets is the basis of single cell studies due to the way the datasets are collected. The advantage with scRNA-seq is that samples are very informative and in healthy bone marrow, often contain a good representation of cells in different stages of the differentiation process. We can therefore assume that we have sufficient information to create a model for the differentiation process using the snapshot data alongside optimal transport frameworks. Building on the assumption that cell states similar to each other have similar gene expression, and that this gene expression can provide information about the order of these cells in a differentiation process, we use a newly proposed optimal transport algorithm to compute probable transitions between cell states as well as a pseudotime ordering of the cells. We use these transitions to extract subsets of cells which represent differentiation trajectories through human bone marrow samples, and order the subset cells in pseudotime. We then estimate expressions of some common marker genes across the subset’s pseudotime in order to verify that the model captures expected biological behaviour. We also apply the method to a sample of cells connected to Acute Myeloid Leukemia (AML) to observe if the changed biological content causes these marker genes to be expressed differently in the sample. The ambition is that this method can aid in further study of the development of AML.}},
author = {{Conradi, Vilhelm and Nelsson, Hugo}},
language = {{eng}},
note = {{Student Paper}},
title = {{Multistage Optimal Transport evaluated on scRNA-seq Human Bone Marrow data}},
year = {{2026}},
}