Multi-omics research proposal and scientific reporting
Practical 26: Genomics, systems biology and multi-omics
Learning objectives
Design a coherent study from question to reporting; specify primary outcomes, unit, sampling and analysis; justify each assay and validation step.
Biological question and data
How would you propose a study of dietary supplementation and later feed efficiency while distinguishing response, prediction and mechanism?
This is a design practical, not an animal-use protocol or a power calculation. The illustrative budget uses arbitrary cost units. Replace these with justified estimates and appropriate approvals in a real project.
No biological download is required. Numerical inputs are hypothetical, not observed animal data. The runnable examples use base R. Count and abundance units are specified where used.
1. Specify the estimand and sampling plan
Write a primary question and separate secondary objectives. Avoid promising all three with one unqualified association analysis.
Show R code
proposal <-data.frame(field=c("primary question","experimental unit","sampling","primary outcome","secondary analysis"),decision=c("diet contrast in later efficiency","individually allocated animal","matched day14 RNA/protein; documented handling","predefined days15–28 efficiency with calibrated units","exploratory profile association; independent prediction evaluation"))proposal
field
1 primary question
2 experimental unit
3 sampling
4 primary outcome
5 secondary analysis
decision
1 diet contrast in later efficiency
2 individually allocated animal
3 matched day14 RNA/protein; documented handling
4 predefined days15<U+2013>28 efficiency with calibrated units
5 exploratory profile association; independent prediction evaluation
The primary contrast, sampling time and outcome window are explicit. Tissue, farm/pen structure, allocation, sample size and feasibility still need specification.
2. Make resource trade-offs visible
Calculate a transparent illustrative assay budget without treating it as a sample-size justification.
Show R code
cost <-c(RNA=3,protein=4,metabolite=2)animals <-24options <-data.frame(design=c("RNA only","RNA plus protein","three layers"),assay_cost=animals*c(cost["RNA"],sum(cost[c("RNA","protein")]),sum(cost)))options
design assay_cost
1 RNA only 72
2 RNA plus protein 168
3 three layers 216
Show R code
stopifnot(options$assay_cost[3]==216)
A cheaper design can support more biological replication or better validation. Cost arithmetic alone cannot identify the statistically or biologically best design.
3. Define the analysis and reporting deliverable
Require each step to connect to the question and assumptions.
Show R code
deliverable <-data.frame(section=c("design","data/QC","analysis","validation","interpretation","governance"),requirement=c("unit, allocation, covariates and sampling","IDs, handling, assay scale, missingness and exclusions","prespecified contrasts and justified integration","independent groups and training-only choices","effect/uncertainty, alternatives and follow-up","access, provenance, sharing and responsible AI"))deliverable
section requirement
1 design unit, allocation, covariates and sampling
2 data/QC IDs, handling, assay scale, missingness and exclusions
3 analysis prespecified contrasts and justified integration
4 validation independent groups and training-only choices
5 interpretation effect/uncertainty, alternatives and follow-up
6 governance access, provenance, sharing and responsible AI
Assess a proposal for coherence: the methods should answer the stated question with measurements available at the right time.
4. Make competing design priorities explicit
A fixed fictional budget can trade assay coverage against independent replication. Compare costed alternatives without assuming the most assays give the most information.
These are arbitrary teaching costs. Real planning requires relevant variance, effect-size sensitivity, allocation structure, missingness and assay feasibility.
5. Write a falsifiable proposal
Specify the primary estimand, phenotype units, target population, allocation unit, sampling schedule, processing design and analysis. For every assay, state the extra uncertainty it addresses. Include a validation strategy and a reporting plan for null or inconsistent results.
A mechanistic proposal should give predicted outcomes under competing explanations. A prediction proposal should define decision-time information and deployment groups. Keep exploratory discovery separate from an independently evaluated procedure. Budget and governance plans should reflect the actual project context rather than the hypothetical numbers here.
Interpretation challenge
How should you justify dropping a proposed omics layer?
Suggested reasoning
Explain whether its measurement is redundant, unavailable at the decision time or unlikely to resolve the primary uncertainty, and what information is lost. The justification follows the scientific target and feasible design.
Interpretation and limits
Formal sample size, actual assay protocols, permits, data agreements, model qualification and collaboration-specific assessment rubrics are outside this invented proposal. The original course assessment weights remain in the curriculum map.
Exercises
Why should a primary outcome be specified?
What is sacrificed when every possible omics layer is added with a fixed budget?
How would a computational-report extension differ from this proposal?
Suggested answers
It anchors design, sample-size reasoning and interpretation instead of selecting the most favourable result.
Potentially replication, assay quality, follow-up or independent validation.
It would implement and evaluate the prespecified pipeline, report exclusions/results/uncertainty and demonstrate reproducibility rather than only promise them.