Research Reporting and Interpretation

1. Methods as a reproducibility record

State data source and access date, sample design, preprocessing, software versions, parameters, reference versions, statistical model, and where scripts are available. Avoid vague phrases such as “standard pipeline.”

Data: public dataset accession and download date
Reference: assembly and annotation release
Software: package versions
Parameters: thresholds, dimensions, seeds
Statistics: model, contrast, correction

2. Figure legends

A legend should define the dataset, groups, visual encodings, preprocessing, statistic, sample count, and abbreviation. It should not make a claim that the figure cannot support.

Figure 1. Mitochondrial QC by condition. Each point is a cell; boxes show median and IQR. Cells were filtered at the pre-specified threshold. n is shown in the panel.

3. Observation versus interpretation

Observation describes what is visible or measured. Interpretation proposes why it may matter. Limitation states what alternative explanations remain.

Observation: treated samples have higher median expression.
Interpretation: treatment may alter the pathway.
Limitation: donors and batch are not fully balanced.

4. Shareable provenance

Publish scripts, environment files, checksums, README, and a license where permitted. Do not publish identifiable human data or credentials.

 git status
 git log --oneline -1
 sha256sum results/figures/qc.webp

5. Reporting Standards in Bioinformatics

Different analysis types have community-agreed reporting standards. Following these ensures your work is reviewable and reproducible.

6. Writing a Methods Section

A complete Methods section should allow another researcher to reproduce your analysis from raw data to final figures. Include:

## Methods

### Data acquisition
RNA-seq reads for 12 tumor/normal pairs were obtained from GEO (GSE123456).
Raw FASTQ files were validated with FastQC v0.12.1 and trimmed with
Trim Galore v0.6.10 (--quality 20 --length 36).

### Alignment and quantification
Reads were aligned to GRCh38 (GENCODE v46) using STAR v2.7.11b
(--outSAMtype BAM SortedByCoordinate). Gene-level counts were obtained
with featureCounts v2.0.6 (-t exon -g gene_id -s 2).

### Differential expression
DESeq2 v1.42.0 was used with default shrinkage (apeglm). Genes with
adjusted p-value < 0.05 and |log2FC| > 1 were considered significant.

### Software environment
All analyses were run on Ubuntu 22.04 with R 4.4.1 and Python 3.11.
A complete conda environment specification is available in the repository.

7. Figure and Table Best Practices

Figures and tables are often the first (and sometimes only) elements reviewers examine. Follow these principles:

• Self-contained: Every figure should be interpretable without reading the full text.
• Labeled axes: Include units, sample sizes, and statistical annotations.
• Color-blind friendly: Use colorblind-safe palettes (e.g., viridis, ColorBrewer).
• Resolution: Minimum 300 DPI for publication; vector formats (PDF, SVG) preferred.

# Example: a publication-quality volcano plot
library(ggplot2)
ggplot(deg_results, aes(x = log2FoldChange, y = -log10(padj))) +
  geom_point(aes(color = significant), size = 1, alpha = 0.6) +
  scale_color_manual(values = c("grey70", "#E41A1C")) +
  labs(
    x = expression(log[2]~"Fold Change"),
    y = expression(-log[10]~"Adjusted p-value"),
    title = "Differential Expression: Tumor vs. Normal",
    subtitle = paste(sum(deg_results$significant), "significant genes (FDR < 0.05, |log2FC| > 1)")
  ) +
  theme_minimal(base_size = 14) +
  geom_hline(yintercept = -log10(0.05), linetype = "dashed", color = "blue")

8. Common Reporting Errors to Avoid

9. Interpretation vs. Over-interpretation

The most common scientific writing error is stating conclusions that exceed the evidence. Follow this hierarchy:

1. Observation: "Gene X shows a 3.2-fold increase in expression (FDR = 0.001) in tumor samples."
2. Interpretation: "This is consistent with the known role of Gene X in cell proliferation pathways."
3. Over-interpretation (avoid): "Gene X causes tumor growth." ← Requires functional validation, not just differential expression.

Practical Exercise

Write a 150-word methods paragraph and figure legend for one plot. Mark each sentence as method, observation, interpretation, or limitation.

Pass criteria: The report includes enough detail to reproduce the plot and does not confuse association with causation or statistical significance with biological importance.

Troubleshooting

If the result is ambiguous, report the ambiguity. Do not change thresholds or omit samples solely to obtain a preferred conclusion.

Next Steps

Continue with Introduction to Bioinformatics and Experimental Design and Batch Effects. Record the software versions, dataset or example inputs, and any decisions you made.