RNA-Seq Differential Expression Dashboard
V2 — True Raw CountsProfessional RNA-Seq analysis using DESeq2 — the gold standard for differential expression.
Dataset: Shemer et al., Immunity 2020 | Comparison: IL10R-Mutant vs Control microglia at 48h post-LPS.
GSE157234
Real published dataset
DESeq2 + apeglm
Modern correct pipeline
Your Data
Upload your own counts
How to use:
- Run analysis_v2.R in RStudio first
- Click Upload Data → Load Demo (V2)
- Explore Volcano, PCA, Heatmap tabs
- Download results from Results Table
Upload Count Matrix
CSV: rows=genes, columns=samples, values= raw integer counts
Upload Metadata
CSV: rows=samples. Must have a condition column with exactly 2 groups.
OR Load V2 Demo Dataset (GSE157234 — True Raw Counts)
Reads pre-computed files saved by analysis_v2.R :
-
results/v2/dds_object_v2.rds -
results/v2/vsd_object_v2.rds -
results/v2/res_df_v2.rds
Run analysis_v2.R first if these don't exist.
Run Analysis (for custom uploads)
Only needed if you uploaded your own count matrix above. Demo data loads with pre-computed results automatically.
Volcano Plot: IL10R-Mutant vs Control (48h post-LPS)
Red=upregulated in Mutant (Tnf, Ccl5, Il6, Il12b, Il1b…). Blue=downregulated in Mutant (P2ry12, Sall1, Tmem119…). Adjust thresholds in sidebar. apeglm LFC shrinkage applied.
PCA: Sample Clustering
Mutant (red triangles) should separate clearly from Control (blue circles) along PC1 — confirms strong hyperactivation effect at 48h post-LPS.
Heatmap: Top DEGs
Z-scored expression. Red=high, Blue=low. Expect Tnf, Ccl5, Il6 high in Mutant; P2ry12, Sall1, Tmem119 low in Mutant.
Differential Expression Results
About This App
RNA-Seq Differential Expression Dashboard — V2
Built with R + Shiny + DESeq2
Dataset
GEO: GSE157234
Paper: Shemer et al., Immunity 53, 1033–1049, 2020
Comparison: IL10R-Mutant vs Control microglia
Timepoint: 48h post-LPS (peak hyperactivation)
Key finding: Without IL-10 signalling, microglia hyperactivate and produce toxic TNF, causing neuronal damage (Figure 3, paper).
V2 Pipeline
- Input: True raw integer counts from Galaxy featureCounts
- Alignment: STAR → mm10 (NCBI RefSeq)
- Quantification: featureCounts (all exons)
- Samples: 6 Control + 3 Mutant (48h post-LPS)
- DESeq2 design: ~ condition
- Filter: genes ≥10 counts in ≥2 samples
- LFC shrinkage: apeglm (modern — replaces deprecated betaPrior)
- Significance: padj<0.05 AND |log2FC|>1
- VST transformation for PCA and heatmap
V1 vs V2 comparison
| V1 | V2 | |
|---|---|---|
| Input | UTAP-normalized | True raw counts |
| Normalization | Double-normalized | Once — by DESeq2 |
| LFC method | Standard | apeglm shrinkage |
| DEGs up/down | 669 / 894 | 621 / 976 |
| Paper (Fig 3E) | 954 / 693 | 954 / 693 |
V2 count differences from paper are due to annotation differences: paper used Gencode vM10 with MARS-seq 3'UTR counting; V2 uses NCBI RefSeq with standard exon counting. Biological conclusions are consistent — key markers confirmed in correct direction.
Developer
Md. Abrar Faiyaj
MSc Biotechnology | Junior Research Collaborator
BRAC University, Dhaka, Bangladesh
This pipeline demonstrates:
- Bulk RNA-seq: SRA → STAR → featureCounts → DESeq2
- Interactive Shiny dashboard development
- Reproducible bioinformatics pipeline design
- Honest limitation documentation
- V1 → V2 methodological improvement
Zenodo DOI: