miRNA Differential Expression and Functional Enrichment of Target Genes
I used R to preprocess expression matrices and model defined contrasts, then connected selected miRNA sets to traceable target evidence and pathway summaries.
Follow the workflow. Select a stage to see it applied to one dataset.
Example run
One dataset · the same workflowExpression matrix
A measured expression value for each miRNA and sample.
- Input
- Expression values + group definitions
- Output
- An analysis-ready matrix
Method notes & example context
Workflow scope. The saved R scripts accept an expression matrix and group sizes. They average repeated identifiers, normalize between arrays, construct a design and contrast matrix, fit limma models, apply empirical Bayes moderation, and export full and filtered result tables. This is matrix-based analysis, not a FASTQ processing pipeline.
Example context. Two independent contrasts were analyzed: 3 days versus control and 14 days versus control, each with three control and three comparison samples. This page illustrates selected rows from the 3-day example only; it does not show a direct 3-day/14-day comparison.
Preprocessing. The input is already on a log scale. Quantile normalization was reproduced with all 1,896 input identifiers; the displayed rows reconcile with the saved fold changes and mean expression to within 3 × 10⁻⁹. Duplicate-ID averaging is part of the script, although these input identifiers are already unique. The statistical model was not refitted for this display.
Contrast analysis. The script uses lmFit → contrasts.fit → eBayes → topTable, then filters by |logFC| > 1 and FDR < 0.05. Full fold changes and FDRs are available on the signal rows. Group definitions use ordered columns and group-size settings, not an automatic sample-sheet parser.
Database hand-off. Target annotation and enrichment are separate database-assisted steps, not a packaged one-click pipeline. This branch is fixed to the archived 3-day upregulated miRNA set; it is not simply the five signals shown in the example.
Target evidence. The displayed miR-344d-3p links to Wnt11, Fzd5 and Pten are miRNet export records with miRTarBase flags and HITS-CLIP evidence (PMID 25083871). These are database records, not experimental validation performed by me. The diagram connects selected records; distances do not encode biological measurements.
Pathway scope. The saved input contains 1,468 target gene IDs, with 616 mapped foreground genes in the KEGG table. Enrichment uses the full target branch, not just the three pictured genes. The bars show selected target-gene counts. Counts and Q values are available on hover, focus or tap. The export's ratio is hits divided by pathway-background genes (pathway coverage), not conventional GeneRatio.