BRM5012 Chap.6 Biomarkers, Drivers and Evidence
Biomarkers, Drivers and Evidence
A biomarker can indicate disease presence or state without causing it, while a driver actively contributes to initiation or progression. This chapter separates those roles and builds an evidence ladder from expression association through replication, specificity, detectability, multi-omics support and functional perturbation.
It also explains why the largest fold change or smallest p-value is not automatically the most important biological candidate.
What this chapter covers
- 01
Biomarker and driver roles
- 02
Effect size and adjusted evidence
- 03
Within-group consistency
- 04
Healthy-tissue specificity
- 05
Clinical detectability
- 06
Independent replication
- 07
Multi-omics support
- 08
Functional perturbation
Worked example · free
Classify two synthetic candidates
- +1Classify Candidate A as a biomarker candidate because it separates states and is practically detectable.
- +1Do not call Candidate A a driver because functional perturbation does not change the phenotype.
- +1Classify Candidate B as having stronger driver evidence because replicated perturbation changes the phenotype.
- +1Keep the different intended uses visible rather than forcing one overall winner.
Key terms
- Biomarker
- A measurable indicator associated with biological state, diagnosis, prognosis or response.
- Biological driver
- A molecule or process that actively contributes to initiation, maintenance or progression of a phenotype.
- Detectability
- The ability to measure a candidate reliably with a realistic assay in the intended specimen.
- Functional perturbation
- An intervention that changes a candidate to test whether the biological phenotype also changes.
Biomarkers, Drivers and Evidence FAQ
Is the largest fold-change gene the best candidate?
Not automatically. Candidate quality also depends on adjusted evidence, consistency, specificity, biological relevance, detectability, independent replication and intended use. A network driver may change modestly, while a dramatic signal may be a downstream consequence.
What evidence supports a biomarker claim?
Show reproducible association with the state, useful performance, specificity for the intended context and practical detectability. Independent validation matters because a strong signal in one dataset can reflect cohort or technical structure rather than general performance.
What additional evidence supports a driver claim?
Add mechanistic reasoning and functional perturbation showing that altering the candidate changes the phenotype. Replication across models and supporting omics or pathway evidence strengthen the claim, but expression association alone does not establish causality.
Assessment move
Build separate checklists for indicator and mechanism claims. For biomarkers, revise effect, adjusted evidence, consistency, specificity, detectability and replication. For drivers, add network position, functional perturbation and replicated phenotypic change. Practise reading candidate tables by intended use rather than by one composite score.
When RNA and protein disagree, write what each layer measured and propose a testable explanation instead of choosing the convenient result. In presentation practice, trace every candidate from the research question through QC, differential analysis and prioritisation criteria. Prepare answers to why this candidate, why this threshold, what independent evidence exists and what experiment would refute the proposed role.
Conclude with the strongest role the evidence supports and make the remaining gap explicit. Start every candidate exercise by choosing an intended use: detection, prognosis, response prediction or intervention. Build a table with separate columns for effect, adjusted evidence, within-group variability, healthy-tissue specificity, assay accessibility, pathway position and independent replication.
Keep these dimensions visible instead of hiding trade-offs in an unexplained total score. Compare a dramatic but variable signal with a smaller, consistent and detectable one, then explain why the best candidate changes with intended use. For a biomarker, practise describing association and performance without implying mechanism.
Ask which specimen would be measured, whether the signal is detectable there, what false positives might arise and whether another cohort reproduces the separation. For a driver, draw the pathway position and state what a perturbation should change if the candidate is causal. Distinguish a knockdown result from expression correlation and require replication of the phenotypic consequence.
When transcript and protein evidence disagree, list timing, translation, degradation and assay coverage as testable alternatives. Do not average the layers into artificial agreement. Trace each proposed label back through design, quality control, differential analysis and enrichment, marking where the claim could fail.
Prepare a rebuttal card for the leading candidate with the strongest competing explanation and the experiment that would discriminate between them. In presentation rehearsal, answer in four moves: intended role, supporting evidence, boundary and next test. The final label should be narrower than or equal to the evidence. A useful marker can be non-causal, and a mechanistic driver can be impractical as a diagnostic assay.
Working through Biomarkers, Drivers and Evidence in BRM5012? Sia is AskSia’s AI Biology tutor — ask any BRM5012 Biomarkers, Drivers and Evidence 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.
