Differential abundance analysis for feature/ observation matrices from platforms such as RNA-seq
This document describes the output produced by the pipeline. The directories listed below will be created in the results directory after the pipeline has finished. All paths are relative to the top-level results directory.
This directory contains the main reporting output of the workflow.
*.html: an HTML report file named according to the value of
params.study_name, containing graphical and tabular summary results for the workflow run.
*.zip: a zip file containing an R markdown file with parameters set and all necessary input files to open and customise the reporting.
Stand-alone graphical outputs are placed in this directory. They may be useful in external reporting, publication preparation etc.
qc/: Directory containing quality control plots from initial processing e.g. DESeq2
exploratory/: Directory containing standalone plots from exploratory analysis. Plots are stored in directories named for the main coloring variable used.
[coloring variable]/png/boxplot.png: Boxplot visualisation of abundance distributions
[coloring variable]/png/density.png: Density visualisation of abundance distributions
[coloring variable]/png/pca2d.png: 2-dimensional PCA plot
[coloring variable]/png/pca3d.png: 3-dimensional PCA plot
[coloring variable]/png/sample_dendrogram.png: A sample clustering dendrogram
[coloring variable]/png/mad_correlation.png: Outlier prediction plots using median absolute deviation (MAD)
differential/: Directory containing standalone plots from differential analysis. Plots are stored in directories named for the associated contrast.
[contrast]/png/volcano.png: Volcano plots of -log(10) p value agains log(2) fold changes
gsea/: Directory containing graphical outputs from GSEA (where enabled). Plots are stored in directories named for the associated contrast.
--study_type maxquant: Directory containing plots produced by the proteus module which is used for processing MaxQuant input. Files are prefixed with the associated contrast and chosen normalization function (if any).
[contrast]/[norm_function].normalized_dendrogram.png: A sample clustering dendrogram after normalization.
[contrast]/[norm_function].normalized_mean_variance_relationship.png: Plots of log intensity vs mean log intensity after normalization of each contrast level.
[contrast]/[norm_function].normalized_distributions.png: A plot of sample distributions after normalization.
[contrast]/raw_distributions.png: A plot of sample distributions without normalization.
Most plots are included in the HTML report (see above), but are also included in static files in this folder to facilitate use in external reporting.
annotation1/: Directory containing annotation matrices generated in the course of analysis
[array platform].annotation.tsv: Annotations derived from an array platform
[GTF name].anno.tsv: Species wise annotations derived from a GTF in RNA-seq analysis
processed_abundance/: Directory containing processed abundance values from initial processing from e.g. DESeq2 or Affy:
[contrast_name].normalised_counts.tsv: Normalised counts table (DESeq2)
[contrast_name].vst.tsv: Normalised counts table with a variance-stabilising transform (DESeq2)
raw.matrix.tsv: RMA background corrected matrix (Affy)
normalised.matrix.tsv: RMA background corrected and normalised intensities matrix (Affy)
differential/: Directory containing tables of differential statistics reported by differential modules such as DESeq2
[contrast_name].deseq2.results.tsv: Results of DESeq2 differential analyis (RNA-seq)
OR [contrast_name].limma.results.tsv: Results of Limma differential analyis (Affymetrix arrays)
gsea/: Directory containing tables of differential gene set analyis from GSEA (where enabled)
[contrast]/[contrast].gsea_report_for_[condition].tsv: A GSEA report table for each side of each contrast
--study_type maxquant: Directory containing abundance values produced by the proteus module which is used for processing MaxQuant input. Files are prefixed with the associated contrast and chosen normalization function (if any).
[contrast]/[norm_function].normalized_proteingroups_tab.tsv: Abundance table after normalization.
[contrast]/raw_proteingroups_tab.tsv: Abundance table without normalization.
differential folder is likely to be the core result set for most users, containing the main tables of differential statistics.
data.rds: serialized R object which can be used to generate a Shiny application
app.R: minimal R script that will source the data object and generate the app
The app must be run in an environment with ShinyNGS installed, or you can see the workflow parameters to deploy to shinyapps.io (see usage documentation).
- Reports generated by Nextflow:
- Reports generated by the pipeline:
pipeline_report*files will only be present if the
--email_on_failparameter’s are used when running the pipeline.
- Reformatted samplesheet files used as input to the pipeline:
- Parameters used by the pipeline run:
- Reports generated by Nextflow:
Nextflow provides excellent functionality for generating various reports relevant to the running and execution of the pipeline. This will allow you to troubleshoot errors with the running of the pipeline, and also provide you with other information such as launch commands, run times and resource usage.
Frequently asked questions
Why are no genes flagged as differentially expressed?
1. Low replication:
Problem: The number of replicates in your RNA-seq experiment may be insufficient to detect statistically significant differential expression.
Suggested course of action: Consider increasing the number of replicates to improve the statistical power of your analysis. Repeating the experiment with greater replication allows for better estimation of biological variation and increases the chances of observing significant differential expression. Consult with experimental design experts or statisticians to determine the appropriate sample size calculation based on your specific research question and resources.
2. Subtle effect:
Problem: The experimental intervention may have a relatively subtle impact on gene expression, making it challenging to detect differential expression using default thresholds.
Suggested course of action: Adjust the analysis parameters to improve sensitivity in capturing smaller changes in gene expression. Try reducing the
differential_min_fold_change parameter to include genes with smaller fold changes. Additionally, consider increasing the
differential_max_qval parameter to relax the significance threshold and capture a broader range of significant p-values or q-values. By fine-tuning these parameters, you increase the likelihood of identifying genes with subtle but biologically relevant changes in expression.
3. Genuinely no differential expression:
Problem: It is possible that the experimental intervention has not significantly impacted gene expression, resulting in the absence of differentially expressed genes.
Suggested course of action: Evaluate the experimental design and the perturbation itself. If the intervention is expected to induce changes in gene expression but no differential expression is observed, revisit the experimental design, biological perturbation, or underlying hypothesis. Consider reassessing the experimental conditions or exploring alternative approaches to investigate other aspects of the biological system.
4. Unaccounted sources of variance:
Problem: Other factors outside the main treatment may introduce variance in gene expression, leading to a decrease in power to detect differential expression.
Suggested course of action: Examine the PCA (Principal Component Analysis) and metadata association plots generated by the workflow. Identify variables associated with components that contribute significantly to the variance in your data. Include these variables as covariates in the contrasts table’s blocking column to account for their effects on gene expression. By incorporating these unaccounted sources of variance into your analysis, you improve the accuracy and power to detect differential expression.
5. Biological complexity and pathway-level effects:
Problem: The experimental intervention may not lead to observable differential expression at the individual gene level, but there may be coordinated changes at the pathway or functional level.
Suggested course of action: Utilize pathway analysis tools such as Gene Set Enrichment Analysis (GSEA), available in this workflow. These tools evaluate the enrichment of gene sets or functional annotations to identify broader biological processes influenced by the experimental intervention. By focusing on pathway-level analysis, you can capture the overall impact of the intervention on biological processes, even if differential expression at the individual gene level is not apparent.
6. Limited options for normalization:
Problem: The nf-core differential abundance workflow currently offers a limited set of normalization methods, which may not fully address the specific normalization requirements of your experiment.
Suggested course of action: If the existing options do not adequately address your experiment’s normalization challenges, consider developing custom normalization modules tailored to your needs. By contributing these modules to the nf-core community, you can expand the range of normalization options available to researchers. Your contributions will help researchers in similar situations and contribute to the continuous improvement and customization of the workflow.
7. Technical variability and batch effects:
Problem: Technical variability and batch effects can introduce noise and confound the detection of differential expression.
Suggested course of action: Address technical variability and batch effects in the experimental design and data analysis. Randomize sample collection, incorporate control samples, and balance samples across different experimental batches. These measures minimize technical variation, enhance the robustness of the analysis, and increase the chances of detecting true differential expression.
8. Workflow issues or bugs:
Problem: Potential issues or bugs in the nf-core differential abundance workflow can affect the detection of differential expression or data analysis.
Suggested course of action: Report any issues or suspected bugs by opening an issue on the nf-core differential abundance workflow repository. Provide specific details, such as software versions, error messages, and relevant data or code snippets. Your feedback is valuable for improving the workflow’s reliability. If you have the technical expertise, consider contributing to the workflow by submitting pull requests to address issues, fix bugs, or propose enhancements.