Gene Expression / Omics

Omics analyses test thousands of hypotheses simultaneously — requiring multiple-testing correction as a non-negotiable disclosure. Licklider enforces Benjamini-Hochberg FDR correction, generates volcano plots with thresholds disclosed, and detects batch effects that must be adjusted for or disclosed as limitations.

STEP 1 — The Scenario

Researcher Profile

A molecular biologist analyzing RNA-seq data from a drug treatment experiment. Tumor cell lines are treated with a targeted compound (n=4 biological replicates) or DMSO vehicle (n=4 biological replicates). Sequencing was performed in two batches: samples 1-4 in Batch 1, samples 5-8 in Batch 2 (2 treated + 2 control per batch).

Research Question

"Which genes are significantly differentially expressed between treated and control cells (adjusted p <0.05, |log2FC| >1)? Are there batch effects that need to be addressed?"

Dataset

• 8 samples: 4 treated, 4 control (2 per condition per batch)
• ~20,000 genes tested simultaneously
• Pre-processed count matrix (gene × sample) provided by the researcher
• Metadata: SampleID, Condition (Treated/Control), Batch (1/2)

STEP 2 — Licklider Analysis Path

1. Data Upload & Metadata Mapping

Licklider accepts the count matrix and the metadata file. Variable & ID Mapping (2.1.2) confirms: Condition is the primary fixed effect, Batch is a potential confounding variable.

2. Batch Effect Detection

Licklider runs Principal Component Analysis (PCA) on the count matrix after variance-stabilizing transformation (VST). If samples cluster strongly by Batch on PC1 or PC2, the Batch / Plate Confounding guard (4.5.3) fires: "Samples appear to cluster by sequencing batch on principal components. Batch effect adjustment is recommended before differential expression testing."

3. Differential Expression Analysis

DESeq2-compatible analysis pipeline: negative binomial model, with Condition as the primary predictor and Batch included as a covariate if a batch effect was detected. For each gene: log2 fold change (log2FC), standard error, Wald statistic, raw p-value, and BH-adjusted p-value (q-value).

4. Multiple Testing Correction

Benjamini-Hochberg (BH) FDR correction is applied across all ~20,000 tests. The significance threshold (q <0.05) and the fold change threshold (|log2FC| >1) are disclosed as the cutoffs used for the volcano plot and the DEG list.

5. Volcano Plot Generation

Volcano plot: log2FC on x-axis, −log10(adjusted p-value) on y-axis. Points colored by significance status: non-significant (gray), significant (adjusted p <0.05 and |log2FC| >1 — red/blue for up/downregulated). Threshold lines shown. Total DEG count annotated (upregulated, downregulated).

6. GO Enrichment (Optional)

Gene Ontology (GO) biological process enrichment analysis on the DEG list, with BH correction applied to the enrichment p-values. Top 10 enriched terms are reported in a bar chart or dot plot.

STEP 3 — Guards & Disclosures Activated

Multiple Comparison Compliance (4.1.5)

BH-FDR correction across all ~20,000 tested genes is mandatory. Reporting raw p-values as the primary significance criterion without correction is blocked. The total number of hypotheses tested (total genes in the count matrix after filtering) must be disclosed.

Batch / Plate Confounding (4.5.3)

If a batch effect is detected on PCA and batch correction is not applied to the statistical model, the guard requires a limitation statement: "Sequencing was performed in two batches. While batch was included as a covariate in the model, residual batch effects cannot be excluded." If batch is completely confounded with condition (all treated in Batch 1, all control in Batch 2), the guard blocks the analysis and requires researcher acknowledgment of the unresolvable confound.

Effect Size + CI Reporting (4.1.2)

log2FC is required alongside the adjusted p-value for every reported DEG. Log2FC without the adjusted p-value, or adjusted p-value without log2FC, cannot be the sole basis for DEG reporting.

Preprocessing Audit Log (1.1.1)

Export review keeps the current preprocessing summary visible, including filtering decisions, missing-data handling where applicable, and related disclosure context. Additional domain-specific omics preprocessing such as normalization or batch correction should be described by the researcher in the final manuscript language rather than assumed to be fully auto-logged by this feature.

N Disclosure (1.2.1)

n=4 biological replicates per condition (not the number of cells sequenced, not the number of reads). The per-sample sequencing depth (library size) is disclosed in the Methods text.

STEP 4 — Export Package

• Figures: PCA plot (batch check), Volcano plot (DEGs), GO enrichment bar chart (optional) — 300 dpi TIFF/PDF
• Statistical Report: DEG table (all genes with log2FC, SE, Wald stat, raw p, BH-adjusted p); DEG count summary (total, upregulated, downregulated); batch detection PCA results; GO enrichment table (if run)
• Methods Text: "Differential expression analysis was performed using DESeq2. Batch was included as a covariate in the negative binomial model. BH-FDR correction was applied across [total genes tested] genes. Significance threshold: adjusted p <0.05 and |log2FC| >1. Library sizes ranged from [X] to [Y] reads. n=4 biological replicates per condition."
• Figure Legend (Volcano): Axis labels, significance thresholds, color coding, total DEG counts, correction method, n
• Preprocessing / provenance support: Gene filtering summary, batch detection context, disclosure notes, and software versions for researcher review