R code and downstream analysis objects for the scRNA-seq atlas of normal and tumorigenic human breast tissue

Abstract

Breast cancer is a common and highly heterogeneous disease. Understanding cellular diversity in the mammary gland and its surrounding micro-environment across different states can provide insight into cancer development in the human breast. Recently, we published a large-scale single-cell RNA expression atlas of the human breast spanning normal, preneoplastic and tumorigenic states. Single-cell expression profiles of nearly 430,000 cells were obtained from 69 distinct surgical tissue specimens from 55 patients. This article extends the study by providing quality filtering thresholds, downstream processed R data objects, complete cell annotation and R code to reproduce all the analyses. Data quality assessment measures are presented and details are provided for all the bioinformatic analyses that produced results described in the study.

Fig. 1

Dataset overview. (a) Diagram showing the data processing pipeline from sample collection to downstream bioinformatics analyses. (b) Schematic overview of the all the integration analyses and the samples involved in each integration analysis. Under each category, the names of the samples are listed and the total number of samples is shown in the bracket.

Fig. 2

Quality control and cell filtering. Box plots of (a) the library sizes and (b) the numbers of detected genes for all the cells in each of the 69 samples before filtering. Boxes show median and quartiles and whiskers show minimum and maximum. Boxes are colored by tumor type. (c) Number of cells in each of the 69 samples. Blue segments show the number of cells that are kept after the cell filtering while red segments show filtered cells. The proportion of filtered cells is labelled on top of each bar.

Fig. 3

Bulk RNA-seq. (a) Number of mapped and unmapped read pairs for each sample in the human mammary gland bulk RNA-seq. (b) MDS plot showing that the bulk RNA-seq samples cluster by cell type. Distances on the plot correspond to root-mean-square log2-fold-change for the top 500 differential genes between each pair of samples. Percentage variance explained is also shown. (c) Biological coefficient of variation for each gene in the bulk RNA-seq data.