Single-cell RNA-Seq deconvolution of glioblastoma bulk transcriptomics
Randelin, Sofia (2020)
Randelin, Sofia
2020
Bioteknologian maisteriohjelma - Master's Degree Programme in Biomedical Technology
Lääketieteen ja terveysteknologian tiedekunta - Faculty of Medicine and Health Technology
This publication is copyrighted. Only for Your own personal use. Commercial use is prohibited.
Hyväksymispäivämäärä
2020-05-08
Julkaisun pysyvä osoite on
https://urn.fi/URN:NBN:fi:tuni-202004284120
https://urn.fi/URN:NBN:fi:tuni-202004284120
Tiivistelmä
Background and aims of study: Glioblastoma multiforme is the most malignant primary brain cancer. Improvement in immunotherapies has shown positive signs in cancer treatment, and brain's immune microenvironment is a promising subject for treatment development. Understanding the qualities and variation between patients would enable classification of tumors and personalised cancer treatment. The aim of the thesis was to characterise cell contents in glioblastoma samples and compare deconvolution to previous results with bulk reference data.
Methods: Tumor microenvironment was analysed with linear regression based gene expression deconvolution. Regression analysis estimated cell content in glioblastoma bulk expression data. Single-cell RNA-seq data was used as reference for cell types. Cell types were identified from single-cell data by marker gene expressions and gene ontology enrichment analysis. Different cluster subsets and averaging methods were used to compare their effects on deconvolution. Results were compared to previous bulk deconvolution study.
Results: Cell type references were found in single-cell analysis. In the single-cell dataset, immune cells were limited to myeloid cells. Different averaging methods had little effect on deconvolution. Although the results was similar with all averaging methods, regression coefficients were highest for vascular and astrocyte reference when 10\% trimmed mean was taken of reference clusters. Single-cell deconvolution had some similarities with bulk deconvolution but the results were not fully comparable due to different reference cell types.
Conclusions: Single-cell reference can be used for estimating cell proportions in bulk transcriptomics. More data is needed for constructing reliable references for multiple cell types. Technical differences between bulk and single-cell data affect the deconvolution. Performance of single-cell reference could be improved with imputation.
Methods: Tumor microenvironment was analysed with linear regression based gene expression deconvolution. Regression analysis estimated cell content in glioblastoma bulk expression data. Single-cell RNA-seq data was used as reference for cell types. Cell types were identified from single-cell data by marker gene expressions and gene ontology enrichment analysis. Different cluster subsets and averaging methods were used to compare their effects on deconvolution. Results were compared to previous bulk deconvolution study.
Results: Cell type references were found in single-cell analysis. In the single-cell dataset, immune cells were limited to myeloid cells. Different averaging methods had little effect on deconvolution. Although the results was similar with all averaging methods, regression coefficients were highest for vascular and astrocyte reference when 10\% trimmed mean was taken of reference clusters. Single-cell deconvolution had some similarities with bulk deconvolution but the results were not fully comparable due to different reference cell types.
Conclusions: Single-cell reference can be used for estimating cell proportions in bulk transcriptomics. More data is needed for constructing reliable references for multiple cell types. Technical differences between bulk and single-cell data affect the deconvolution. Performance of single-cell reference could be improved with imputation.