BRM5012 Chap.4 Differential Expression Analysis
Differential Expression Analysis
Differential expression analysis estimates how gene expression changes between named conditions while accounting for variability and many simultaneous tests. This chapter keeps comparison direction attached to every sign, converts log2 fold change into conventional ratios, explains adjusted p-values and distinguishes volcano, mean-difference and set-membership plots.
It finishes by turning gene-level results into bounded candidate claims rather than causal declarations.
What this chapter covers
- 01
Named comparison direction
- 02
Log2 fold change
- 03
Conventional fold change
- 04
Average expression
- 05
Adjusted p-values
- 06
Volcano plots
- 07
Mean-difference plots
- 08
Candidate interpretation
Worked example · free
Interpret two synthetic fold changes
- +1Write treated versus control as the declared comparison.
- +1Calculate two raised to 2, giving a four-fold increase for Gene R.
- +1Calculate two raised to minus 1.5, giving approximately 0.354 for Gene S.
- +1Express Gene S as treated at 35.4% of control, or control approximately 2.82 times treated.
- +1Keep adjusted evidence, abundance and consistency separate from effect magnitude.
Key terms
- Differential expression
- A model-based estimate of expression change between named conditions under the study design.
- Log2 fold change
- A symmetric logarithmic expression ratio whose sign reverses when the comparison direction is reversed.
- Volcano plot
- A plot combining effect direction and magnitude with statistical evidence for many tested genes.
- Mean-difference plot
- A plot showing expression change against average expression without displaying adjusted p-values directly.
Differential Expression Analysis FAQ
Why does comparison direction matter?
A positive log fold change means higher expression in the numerator condition relative to the reference. Reverse the contrast and the sign reverses while the data remain unchanged. Keep the contrast visible beside tables, plots and narrative.
Why are adjusted p-values used?
High-throughput analysis tests many genes at once, so small raw p-values can occur by chance. Adjustment controls the chosen family-wide error criterion. Report the applied threshold and do not substitute raw values merely to retain a preferred gene.
Which plot should show abundance-dependent change?
Use a mean-difference plot because it places change against average expression. Use a volcano plot when the central question combines direction and statistical evidence. The mean-difference plot does not show adjusted p-values, so report significance separately.
Assessment move
Put the comparison direction at the top of every practice page and refuse to interpret a sign without it. Recompute powers of two for a range of positive and negative log2 fold changes until you can express both the direct ratio and the reciprocal clearly. Build a plot-choice table: volcano for effect and evidence, mean-difference for effect and abundance, and Venn diagrams for selected-set overlap.
Trace one gene through the results table, plot label and written conclusion to catch sign or naming drift. Practise separating effect size, adjusted evidence, average abundance, consistency and biological relevance. Finish every answer with the strongest safe claim: differential association, candidate prioritisation or pathway hypothesis. Reserve biomarker and driver language for the additional evidence those roles require.
Create contrast cards with a numerator, reference and plain-language question. For every card, predict the sign of a gene that is higher in each group before looking at a result table. Reverse the contrast and verify that signs reverse while adjusted evidence and the underlying samples do not.
Use a small conversion table for log2 changes of negative two, negative one, zero, one and two, then practise wording values below one with a reciprocal so the direction remains clear. Audit a synthetic result through four surfaces: original table, volcano plot, mean–difference plot and final paragraph. Preserve identifier, sign, comparison and selection rule at every transfer.
Define the word top before sorting; largest positive change, largest absolute change, smallest adjusted p-value and largest Wald statistic serve different purposes. Add average abundance to stop a visually dramatic low-count estimate from becoming an unsupported headline.
For multiple testing, explain why thousands of simultaneous tests make a raw threshold inadequate and why an adjusted value cannot replace effect size or consistency. Review mappings from transcript identifiers to gene symbols and count how many records are lost or duplicated. When a gene is absent from a shortlist, ask whether it failed measurement, filtering, annotation or the stated criterion.
Build conclusions in layers: model-supported difference, pathway hypothesis, candidate for validation, then any stronger role only when external evidence exists. Finish each exercise by proposing one check that could overturn the preferred explanation. This routine makes arithmetic, plots and biological language part of one auditable evidence chain.
Working through Differential Expression Analysis in BRM5012? Sia is AskSia’s AI Biology tutor — ask any BRM5012 Differential Expression Analysis question and get a clear, step-by-step explanation grounded in how BRM5012 is taught and assessed. Read this chapter free, then take your hardest questions to Sia.
