Sleuth
The DE Quick Start tutorial provides an introduction to basic DE functionality and navigation.
Rationale and background:
Sleuth is a tool for the analysis and comparison of multiple related RNA-Seq experiments. Key features include:
- The ability to perform both transcript-level and gene-level analysis.
- Compatibility with Kallisto enabling a fast and accurate workflow from reads to results.
- The use of bootstraps to ascertain and correct for technical variation in experiments.
- An interactive app for exploratory data analysis.
To use sleuth, RNA-Seq data must first be quantified with Kallisto, which is a program for very fast RNA-Seq quantification based on pseudo-alignment. An important feature of Kallisto is that it outputs bootstraps along with the estimates of transcript abundances. These can serve as proxies for technical replicates, allowing for the ascertainment of the variability in estimates due to the random processes underlying RNA-Seq as well as the statistical procedure of read assignment. Kallisto can quantify 30 million human reads in less than 3 minutes on a Mac desktop computer using only the read sequences and a transcriptome index that itself takes less than 10 minutes to build. sleuth has been designed to work seamlessly and efficiently with Kallisto, and therefore RNA-Seq analysis with Kallisto and sleuth is tractable on a laptop computer in a matter of minutes. More details about Kallisto and sleuth are provided the papers describing the methods:
- Nicolas L Bray, Harold Pimentel, Páll Melsted and Lior Pachter, Near-optimal probabilistic RNA-seq quantification, Nature Biotechnology 34, 525–527 (2016), doi:10.1038/nbt.3519
- Harold Pimentel, Nicolas L Bray, Suzette Puente, Páll Melsted and Lior Pachter, Differential analysis of RNA-seq incorporating quantification uncertainty, in press.
Sleuth has been designed to facilitate the exploration of RNA-Seq data by utilizing the Shiny web application framework by RStudio. The worked example below illustrates how to load data into sleuth and how to open Shiny plots for exploratory data analysis. The code underlying all plots is available via the Shiny interface so that analyses can be fully “open source”.
Prerequisites
A CyVerse account. (Register for a CyVerse account here - user.cyverse.org.)
Input
experiment design matrix file: Matrix file describing the experiment check the sleuth documentation https://pachterlab.github.io/sleuth_walkthroughs/trapnell/analysis.html. Make sure the sample name matches the Kallisto output sample folder names and in the design matrix file the sample column is represented by column name "sample".
- fullmodel formula: e.g ~condition. Sleuth applies the likelihood ratio test—where the full model contains labels for the samples
- reduced model formula: e.g ~1 for the intercept model. Sleuth applies the likelihood ratio test—where the reduced model ignores labels, intercept model is given by ~1
- Experimental covariate used in Sleuth plots: This is usually the column name in the design matrix file for which differential experiment is performed. e.g group or condition etc
- Directory of bootstrapped Kallisto output abundance estimates: directory of Kallisto quantifications output files
Test/sample data:
The test data are provided for testing trim_galore-0.4.1 is in here - /iplant/home/shared/iplantcollaborative/example_data/Kallisto_Sleuth:
Use the following inputs/outputs and parameters for testing trim_galore-0.4.1
Input
experiment design matrix file: design_matrix
- fullmodel formula: ~condition
- reduced model formula: ~1
- Directory of bootstrapped Kallisto output abundance estimates: kallisto_quant_output
Output Reports:
After successful completion of the run, expect the following files as output:
- density.jpg
- mean_variance.jpg
- pca.jpg
- results_trans_sig.tsv
- results_trans.tsv
- sleuth_obj
For more detailed information about these outputs, please refer to this link - https://pachterlab.github.io/sleuth_walkthroughs/trapnell/analysis.html
More information about Scythe-0.991 can be found at trim-galore manual.