Welcome to SAMAR (Speedy, Assembly-free Method to Analyze RNA-seq expression data) -- a quick-and-easy way to perform differential expression (DE) analysis in non-model organisms.
For non-model organisms, conventional DE analysis begins by constructing a de-novo transcriptome assembly and annotating it against a high-confidence protein database -- the assembly serving as a reference for read mapping and the annotation allowing for functional analysis of DE genes. Assembly can be complicated, error-prone, requires massive computational resources, and time-consuming. This pipeline provides a shortcut in which RNA-seq reads are directly aligned to the reference proteome which would have been otherwise used for annotation in the assembly-based approach. This approach is super fast and light-weight.
Specifically, SAMAR performs the following steps:
- Uses LAST to learn the alignment scoring parameters suitable for the input data, and to estimate the paired-end fragment size distribution of paired-end reads,
- Uses LAST to directly align RNA-seq reads to the high-confidence proteome that would have been otherwise used for annotation, and generates counts,
- Estimates the counts of reads aligning to each protein entry in the reference,
- Uses tximport to read in the counts (and optionally to aggregate isoform-level counts into gene-level counts),
- Runs DESeq2 for differential expression analysis based on the count data.
This pipeline requires the package manager Conda and the workflow management system Snakemake. All other dependencies are handled automatically by Snakemake.
Conda can be downloaded and installed from Miniforge. Once installation is complete, you can test your Conda installation by running:
$ conda list
Snakemake recommends installation via Conda:
$ conda create -c conda-forge -c bioconda -n snakemake snakemake=8.24.1
This creates an isolated enviroment containing the latest Snakemake. To activate it:
$ conda activate snakemake
To test snakemake installation
$ snakemake --help
Download SAMAR from the online repository, or using the command line:
git clone https://github.com/bioinfodlsu/samar.git
cd samar
A toy example is provided in the test_data folder.
In this example, there are 6 RNA-seq samples, of which 3 are the "control" group replicates and 3 are the "treated" group replicates.
The paths to the files and information about the experiment design is specified in config_test.yaml.
From the top-level directory of SAMAR_lite, run:
$ snakemake -p --configfile testdata/config_test.yaml --use-conda --cores all
Warning: This might take some time (5-10 mins), since for the first run, Snakemake needs to install all dependencies prior to running the actual computations.
Once the run is complete, you can check if some outputs are generated in the newly created test_output folder.
The contents of the output folder are described in the section below.
The pipeline requires, at the very least: (1) RNA-seq fastq (or fasta) reads and (2) a reference protein database in fasta format.
These and other input parameters are specified via a YAML-format config file -- a template config_template.yaml is provided in the config folder.
Section~3 below provides a list of all parameters that can be defined in the config file.
After constructing a config.yaml file and with the snakemake conda environment you created earlier activated, you can call the pipeline from the top-level directory of SAMAR:
cd samar
snakemake --configfile <my_config.yaml> --use-conda -p --cores all
Please see 2.4 below for a simple example.
Output is stored inside the folder specified in the out_dir entry of the config file. Inside it, the alignments can be found in the last_alignments folder, the count data in the counts folder, and the results of differential expression analysis in the DEanalysis folder.
Inside the DEanalysis folder in test_output, you will find the following files:
- DE_count_summary.txt, which contains the count of up- and downregulated genes
- Test_result-Group_treated_vs_control.csv, which contains the table of the results of hypothesis testing for each gene in the reference.
- DESeq2_fit.RDS which can be loaded using R, if further analysis is required.
For the example run above, the first file reveals that only 1 gene is differentially expressed; while the second file allows the identification of this gene to be UPI0000000AE2 which is upregulated (log2FoldChange > 0 and padj < 0.01).
In general, the number of hypothesis tests depends on the number of factors, the number of levels per factor, and the design (i.e. the factorial model that is hypothesized to cause a difference in gene expression). These have to be specified in the config file using the keywords: factors , sample_info, and design, respectively.
Please note, especially for complicated designs, that the first sample in sample_info will be the reference group, and that significance testing is only done for the difference of the other levels (or interaction of these levels) vs this reference group. For advanced DE analysis like testing contrasts, a DESeq2 object is available in the loadable R data to perform such functions.
In case you wish to perform your own differential gene expression analysis, the counts folder contains a .counts table for each sample, where the last column (NumReads) should be used for differential expression analysis.
Apart from the paths to the read file and the reference proteome, the following parameters can be set in the YAML config file.
| Keyword | Possible values | Default | Description |
|---|---|---|---|
| filetype | fasta, fastq | fastq | - |
| training | yes, no | yes | Train the alignment scoring parameter for the input data. The default scoring scheme is BLOSUM62 |
| deanalysis | yes, no | yes | If set to "no", the pipeline does not proceed to DE anlaysis and halts after counting. If keyword is not provided, value defaults to "yes" |
| factor | [factor1, factor2, .. ] | - | Name of factors in the study |
| sample_info: sample_i: | [level of factor1, level of factor2, ..] | - | Describes the factor levels which the sample corresponds to |
Assembly-free rapid differential gene expression analysis in non-model organisms using DNA-protein alignment, BMC Genomics 23, 97,(2022) Link.
For comments, suggestions, issues, please send an email to anish.shrestha--atmark--dlsu.edu.ph