Logo image
Standardizing RNA-seq data comparison and unambiguous cell type classification
Thesis   Open access

Standardizing RNA-seq data comparison and unambiguous cell type classification

Roze Alzabey
Master of Science (M.S.), Drexel University
Jun 2021
DOI:
https://doi.org/10.17918/00000418
pdf
Alzabey_Roze_20213.97 MBDownloadView

Abstract

Engineering
RNA sequencing (RNA-seq) has been established as a high throughput sequencing method that provides gene expression profiling. However, with the vast amount of data generated through RNA-seq, it remains a challenge to extract meaningful interpretations of the data. The workflow of RNA-seq includes multiple steps starting with experiment design, followed by RNA preparation, sequencing library preparation, sequencing and data analysis. Variability in sequencing data occurs due to differences in sequencing protocols at various levels of the process, and a lack of standardized methods have made it very difficult to compare across different datasets, particularly those generated in separate laboratories. In this project, we ask the question of how well RNA-seq data of the same cell type compare across two different labs utilizing different methods in their RNA-seq protocols. We first investigate reproducibility of the data by comparing replicates within each dataset using correlation analysis. We then investigate variability across datasets by establishing a standardized analysis to compare across RNA-seq libraries. Using this analysis, we highlight variability that emerges from selected methods of the library preparation process. Next, using an open-source RNA-seq dataset that is currently being used as a reference to classify cell types following single cell RNA-seq (scRNA-seq), we investigate how methodological variability leads to limitations in unambiguous cell type classification. We perform this work using the model organism Drosophila melanogaster, due to the availability of advanced genetic tools and its well-established anatomy.

Metrics

48 File views/ downloads
40 Record Views

Details

Logo image