This repository contains supporting code to facilitate reproducible analysis. For details see the biorxiv preprint. If you find bugs please create a github issue.
The quasi-UMI normalization method is under development as a standalone R package.
The purpose of this method is to remove PCR distortion from scRNA-seq read counts by normalizing to quasi-UMIs (QUMIs). QUMIs approximate the true (unmeasured) UMI counts. Once read counts are transformed to QUMIs, the count matrix can be passed to UMI-specific methods for feature selection and dimension reduction.
Will Townes and Rafael Irizarry
Implementation of quasi-UMI normalization methods, auxiliary functions, and methods used in comparisons.
- existing.R - wrapper functions for PCA, tSNE, ZINB-WAVE, etc.
- nblomax.R - density, visualization, and maximum likelihood estimation of compound negative binomial- Lomax families. These are discrete distributions with heavy power-law tails.
- poilog.R - density, visualization, and maximum likelihood estimation of compound Poisson-lognormal families. These are discrete distributions with heavy tails (but not as heavy as power law).
- quminorm.R - quasi-UMI normalization. The most important function is quminorm_matrix which applies the normalization to a matrix of read counts with features in rows, samples in columns. Also of interest is lpmf which visualizes heavy-tailed data with log-log plots (histograms with logarithmic binning).
Code for making graphs that involved multiple datasets.
Analysis of various real scRNA-seq datasets. The Rmarkdown files can be used to produce figures in the manuscript. The util subfolder contains numerous scripts for generating count matrices from FASTQ files, working with BUS files, and running QUMI normalization on large matrices using parallelization on a high performance computing cluster.
Miscellaneous utility functions.