Structure of a Scientific Paper (IMRaD)

Every peer-reviewed paper in biology, genomics, and bioinformatics follows the same skeleton. Learn it once — and you will be able to read any paper, write any paper, and review any paper with confidence.

✦ Lesson 1 of 12 ⏱ 45 min 📖 Beginner 🆓 Free

01 What is IMRaD?

🧬 Why this matters for you

Every paper you will ever read in Nature Plants, Bioinformatics, PLOS Genetics, or any other peer-reviewed journal uses the same four-part structure: Introduction, Methods, Results, and Discussion — abbreviated as IMRaD.

When you understand why each section exists and what belongs in it, you can read a 12-page paper in 20 minutes by knowing exactly where to look. And when you write your own papers — including your MSc thesis chapters — you will never face a blank page again, because you always know which section you are filling in.

IMRaD is not just a formatting convention. It is a logical argument structure. Each section answers one fundamental question that a sceptical reader would ask:

I
Introduction
Why did you do this study? What problem are you solving?
M
Methods
How did you do it? Could someone else repeat this?
R
Results
What did you find? Just the facts — no interpretation.
D
Discussion
What does it mean? How does it fit the bigger picture?

Notice that Results and Discussion are deliberately separated. You are not allowed to interpret your data in the Results section — that separation is one of the most important discipline-building rules in science writing.

💡
Historical note: IMRaD was not invented overnight. Before the 1950s, many papers had no standard structure. The American National Standards Institute (ANSI) formalised IMRaD in 1979. Today, over 95% of life-science journals require it.

02 Why IMRaD exists

IMRaD solves a problem that was crippling science communication in the early 20th century: papers were written like detective novels, with the reader following the author's thought process from start to finish. That made papers long, hard to skim, and difficult to compare.

Modern science produces thousands of papers per week. A researcher reading literature on RNA-seq normalisation methods cannot read every paper from front to back. IMRaD lets them:

  • Scan the Abstract to decide if the paper is relevant.
  • Jump to Methods to check if the experimental design is sound.
  • Read only the Results figures to understand the main finding.
  • Check Discussion to see if the authors' interpretation matches the data.

For bioinformaticians especially, the Methods section is critical — it tells you exactly which tools, parameters, reference genomes, and statistical thresholds were used, so you can reproduce or adapt the analysis.

🌱 Bioinformatics context

When you publish your genomic selection BPV study, editors at journals like Theoretical and Applied Genetics or Plant Genome will expect perfect IMRaD structure. Your Methods will need to specify: the rrBLUP and BGLR R packages used, the LD-pruning thresholds, the cross-validation scheme, and the BPV weighting formula. A clear Methods section is what allows other sorghum breeders to replicate your genomic selection pipeline.

03 The hourglass shape of a paper

The best mental model for IMRaD is an hourglass. A good scientific paper starts broad, narrows to your specific study, and then expands back out to the implications for the field.

INTRODUCTION broad → narrow M + R DISCUSSION narrow → broad
Introduction (broad → narrow): Start with the global problem (e.g. food security, crop improvement), narrow to the specific gap in knowledge your study addresses, and end with your specific research question or hypothesis.
Methods & Results (narrow): The waist of the hourglass. This is your specific study — your exact samples, tools, algorithms, and findings. No broad claims here.
Discussion (narrow → broad): Start by interpreting your specific results, then connect them to the wider field, compare with existing literature, acknowledge limitations, and end with future directions.
⚠️
Common mistake: Many students write a Discussion that simply repeats their Results in different words. A Discussion should interpret results, not re-describe them. "The SNP at chromosome 4 was significantly associated with yield" belongs in Results. "This QTL region overlaps with a previously reported drought-tolerance locus in maize, suggesting a conserved mechanism" belongs in Discussion.

04 The Introduction section

The Introduction has one job: convince the reader that your study was necessary. It does this in three moves, sometimes called the Create a Research Space (CARS) model, introduced by linguist John Swales:

Move What you write Example (genomics)
Move 1
Establish the territory
Why does this research area matter? What is already known? Cite key references. "Genomic selection has transformed plant breeding by enabling prediction of complex traits without phenotyping every individual (Meuwissen et al., 2001)."
Move 2
Identify the gap
What is missing, inconsistent, or unknown? This is your justification. "However, optimal weighting of Best Progeny Value (BPV) in recurrent selection cycles for orphan crops such as Sorghum bicolor remains poorly characterised."
Move 3
Occupy the niche
State your objective, hypothesis, or what your study does to fill the gap. "Here we compare five BPV weighting strategies using rrBLUP and BGLR models across three sorghum populations to identify the optimal recurrent selection scheme."
💡
Length: An Introduction for a research article is typically 400–700 words (roughly 3–5 paragraphs). It is not a literature review — it is a focused argument for why your study was needed. Every sentence should serve one of the three CARS moves.

05 The Methods section

The Methods section answers one question: could a competent researcher in your field reproduce your study from this description alone? If yes, your Methods are complete.

For bioinformatics papers, Methods typically include:

  • Study system / data source — species, accession numbers (e.g. NCBI SRA), population size, environmental conditions.
  • Wet lab protocols (if applicable) — DNA extraction method, library preparation kit, sequencing platform.
  • Bioinformatics pipeline — every software tool with version number and key parameters. "We used STAR v2.7.10a for alignment with default parameters" is not enough — list the non-default parameters you changed.
  • Reference genome / databases — assembly version, annotation version, database release date.
  • Statistical analysis — model used, significance threshold, multiple testing correction method, cross-validation scheme.
  • Code and data availability — GitHub link, Zenodo DOI, or supplementary data statement.
🔬 Real example from bioinformatics

A Methods paragraph for a DESeq2 differential expression analysis might read:

"Raw read counts were normalised using DESeq2 v1.38.0 (Love et al., 2014) in R v4.2.1. Genes with fewer than 10 reads across all samples were excluded. Differential expression was tested using the negative binomial Wald test, with an adjusted p-value threshold of 0.05 (Benjamini-Hochberg correction) and a minimum log₂ fold-change of 1.5."

Notice: tool name, version, citation, filtering threshold, statistical test, significance threshold, correction method, and effect size threshold are all specified.

⚠️
Version numbers are not optional. Bioinformatics tools change between versions. A reader trying to reproduce your analysis needs to know exactly which version of GATK, Samtools, or STAR you used. This is also required for reproducibility under FAIR data principles.

06 The Results section

The Results section presents your findings — without interpretation. Think of it as a guided tour of your figures and tables. You tell the reader what each figure shows, what the key numbers are, and whether results were statistically significant. You do not say what those findings mean.

Correct (Results)Wrong (belongs in Discussion)
"BPV weighting strategy 3 produced the highest genomic prediction accuracy (r = 0.72, p < 0.001) across all three populations." "This superior accuracy suggests that strategy 3 is the most effective for practical sorghum breeding programmes." ❌
"Of 18,432 SNPs genotyped, 12,847 passed quality filters (MAF > 0.05, missing rate < 10%)." "The high SNP retention rate confirms the quality of our genotyping data." ❌
"Figure 2 shows the Manhattan plot for grain yield GWAS, with a significant peak at chromosome 4 (−log₁₀P = 8.3)." "This QTL on chromosome 4 is likely involved in drought tolerance." ❌
💡
Structure tip: Organise Results by your research questions or hypotheses, not by the chronological order in which you did the experiments. Each paragraph typically introduces a figure or table, states the key finding, and gives the key statistic.

07 The Discussion section

The Discussion is where you think out loud — scientifically. It has a conventional structure of its own:

Sub-sectionWhat it contains
Opening statement One or two sentences restating your main finding — but this time as a conclusion, not a raw result. "This study demonstrates that…"
Interpretation What do your findings mean biologically or methodologically? Connect results to mechanisms.
Comparison with literature How do your findings agree or disagree with previous work? Explain unexpected results.
Limitations What are the weaknesses of your study? Small sample size? Limited geographic scope? Missing data?
Future directions What should the next study do to build on or address the limitations of your work?
Conclusion A short, strong closing statement about the overall significance of your work. This is often a single paragraph or is its own section.
✍️ Writing the Discussion for your BPV thesis

When you write the Discussion for your genomic selection thesis, the comparison-with-literature sub-section is where you cite studies on BPV in maize (e.g., Witcombe et al., 2013) or sorghum. If your results agree, explain why. If they disagree, explain the most likely reason — different environment, different population structure, different marker density.

Your limitations section might mention that your study used only one growing season or that the training population was small (common in orphan crop studies). This is not a weakness of your thesis — acknowledging limitations shows scientific maturity.

08 Title, Abstract & References

IMRaD describes the body of the paper. But three other components are equally important:

ComponentPositionKey rule
Title Before everything Should contain the main finding or the key variables studied. Specific > general. "Genomic prediction of grain yield in sorghum using rrBLUP and BGLR models" is better than "A study on genomic selection".
Abstract Before Introduction A structured mini-version of the whole paper: background → objective → methods (brief) → key results → conclusion. Usually 150–300 words. Written last.
Keywords After Abstract 5–8 terms that indexing databases (PubMed, Scopus) will use to classify your paper. Choose terms your target readers would search.
References After Discussion Formatted in the journal's required style (APA, Vancouver, author-year, numbered). Always use a reference manager — covered in Lesson 8.
💡
Write the Abstract last. The Abstract summarises a paper that exists. If you write it first, you will rewrite it six times. Write it after all other sections are complete.

09 A real journal paper, mapped to IMRaD

Let us walk through a typical paper in plant genomics — your own published paper is a perfect case study. The Hossain et al. (2015) paper in the Journal of Rice Research follows this structure:

📄 Hossain et al. (2015) — IMRaD mapping

Title: Specifies the crop (rice), the trait studied, and the experimental approach.

Introduction: Establishes the importance of the rice trait studied → identifies what previous approaches could not determine → states the objective of your study.

Methods: Describes the plant material, growing conditions, measurements taken, and statistical analysis used.

Results: Presents the descriptive statistics, comparisons between varieties, and any significant differences found — with tables and figures. No interpretation.

Discussion: Interprets the variety differences in the context of breeding value, compares with earlier literature, notes limitations, and suggests which varieties should be prioritised.

This is exactly the structure your thesis chapters and future publications will follow.

When you read your next paper, try to identify these sections explicitly. Ask yourself: Is this sentence describing a finding or interpreting it? That question will train your ability to write each section correctly.

Advertisement Sponsored content

10 Exercises

1
IMRaD classification

Read each sentence below and decide which IMRaD section it belongs to (Introduction, Methods, Results, or Discussion). Write your answer before checking.

A. "RNA was extracted using TRIzol reagent (Invitrogen) according to the manufacturer's protocol."

B. "Despite the high prediction accuracy observed here, the small training population (n = 120) may limit generalisability to other environments."

C. "Differential expression analysis identified 847 upregulated and 312 downregulated genes (adjusted p < 0.05) in drought-stressed plants."

D. "Understanding the molecular basis of drought tolerance in sorghum is critical for breeding climate-resilient varieties."

▶ Show answers

A → Methods. A specific protocol with brand name and instruction source. This is exactly the level of detail Methods requires.

B → Discussion (Limitations sub-section). The sentence acknowledges a weakness ("may limit generalisability") — this is interpretation and belongs in Discussion.

C → Results. Numbers, a statistical threshold, but no interpretation of what the genes do or why they changed. Pure findings.

D → Introduction (Move 1: Establish the territory). A broad statement about why the research area matters — the opening move of a CARS Introduction.

2
Write a one-paragraph Introduction

Using the three CARS moves (establish territory → identify gap → occupy niche), write a short Introduction paragraph for an imaginary study:

Scenario: You are publishing your MSc thesis on BPV weighting in recurrent genomic selection for sorghum grain yield using rrBLUP and BGLR.

Your paragraph should be 80–120 words and contain all three moves. It does not need to be perfect — focus on structure.

▶ Show a model answer

Move 1 (territory): "Genomic selection (GS) has revolutionised plant breeding by enabling prediction of complex traits from genome-wide marker data (Meuwissen et al., 2001). In orphan crops such as sorghum (Sorghum bicolor), where phenotyping resources are limited, GS offers a cost-effective path to genetic gain."

Move 2 (gap): "However, recurrent GS schemes in sorghum require optimal weighting of Best Progeny Value (BPV) to maximise long-term genetic gain without excessive loss of genetic diversity. This weighting problem remains poorly characterised for sorghum."

Move 3 (niche): "Here we evaluate five BPV weighting strategies using rrBLUP and BGLR genomic models across three sorghum breeding populations to identify the most effective recurrent selection scheme."

3
Find an open-access paper and map it

Go to PubMed or bioRxiv and find any open-access paper in plant genomics or bioinformatics. Read it and answer these questions:

  • What are the three CARS moves in the Introduction?
  • Does the Results section contain any interpretation, or is it purely descriptive?
  • Does the Discussion open with a restatement of the main finding?
  • Does the Methods section give version numbers for all software used?

This exercise takes 20–30 minutes and is the single most effective way to internalise IMRaD structure.

You have covered IMRaD structure, the hourglass model, CARS moves, and the role of each section. Ready to mark this complete?

Advertisement Your ad here