This repo is associated with the publication:

Zhiyuan Hu, Mara Artibani, Abdulkhaliq Alsaadi, ..., Tatjana Sauka-Spengler, Christopher Yau *, Ahmed A. Ahmed *. (2020). The repertoire of serous ovarian cancer non-genetic heterogeneity revealed by single-cell sequencing of normal fallopian tube epithelial cells. Cancer Cell. Volume 37, Issue 2: p226-242. doi: https://doi.org/10.1016/j.ccell.2020.01.003

• Rmd is the code file
• html is the Rmarkdown report

There are four file folders corresponding to different parts in the manuscript.

We used differential expression (DE) analysis and pseudotime analysis to compare the freshly dissociated cells and cultured cells.

• In culture_effect/culturing_180918.Rmd and culture_effect/M_S4_0913_culturing.html we used DE analysis and pathway analysis to investigate the difference between cells from various sources.

• In culture_effect/PhenoPath_181010.Rmd and culture_effect/M_S4_1010_PhenoPath.html we used psuedotime ananlysis (PhenoPath, Campbell and Yau) to dig deeper.

Before we entering the "true" analysis, we must do some QC steps to avoid the inclusion of tumour cells into our analysis. A key characteristics of HGSOC cells is the frequent copy number variants (CNVs), which is similar to the glioblastoma cells.

• In cnvQC/HoneyBadger_fresh_secretory_exprs20180706.Rmd and its report cnvQC/HoneyBadger_fresh_secretory_exprs20180706.html you can see how we use HoneyBadger (Fan et al., 2018) to infer the CNV from cells dissociated from FT of cancer patients.

• In cnvQC/P11528_tumour_FTE_SNPsCNVs20180711.Rmd and its report cnvQC/P11528_tumour_FTE_SNPsCNVs20180711.html, we revealed some results that were not included in the manuscript. By comparing the SNVs called from the scRNA-seq data and the ones called from WES data, we found that the cells from pt11528 carrying CNVs also harhoured the pathological p53 mutation, indicating that they are either early lesion or metastasis.

The part contains the some key coding for the manuscript.

• In clustering/Github_clusteing_all_data.Rmd and its report clustering/Github_clusteing_all_data.html, we first clustered all the FT cells from cancer patients, identifying major FTE cell types.

• In clustering/Github_manuscript_clustering.Rmd and its report clustering/Github_manuscript_clustering.html, we further clustered the secretory cells into fine-grained subtypes.

• In clustering/visualisation_secretory.Rmd and its report clustering/visualisation_secretory.html, you will see how the plots were produced for the manuscript.

• In clustering/data_integration.Rmd and its report clustering/data_integration.html, we used Seurat v3 to integrate the secretory cells from cancer patients and from benign donors, in which the existence of the secretory subtypes was valdiated.

In the last part, we used the information obtained by scRNA-seq to deconvolute TCGA, AOCS and other datasets from CuratedOvarianData.

• In deconvolution/deconvolution_analysis.Rmd and its report deconvolution/deconvolution_analysis.html, we performed deconvolution and survival analysis. The deconvolution was conducted by using Cibersort (Newman et al.).

• In deconvolution/DEanalysis_EMThigh_TCGA.Rmd and its report deconvolution/d DEanalysis_EMThigh_TCGA.html, we studied the molecular characteristics of those EMT-high tumours that had worse prognosis.