Methodology

Analysis methodology and reproducibility standards

SBM Lab's analysis guides follow the principles of reproducible science. This page sets out the methodological standards we apply when writing content.

1. Reproducibility Principles

Every analysis guide meets the following four reproducibility criteria:

  • Tool versions stated — the exact version of each R/Python package used (e.g. DESeq2 v1.42.0)
  • Dependency management — system requirements, operating system and dependent libraries are listed
  • Parameters disclosed — every analysis parameter is stated in the text (including when defaults are used)
  • Runnable code — code you can copy, paste and actually run

2. Tool Version Policy

Tool version baselines by field (as of May 2026):

ToolRecommended versionNotes
DESeq2v1.42+Bioconductor 3.18+
edgeRv4.0+Bioconductor 3.18+
limmav3.58+
Seuratv5.0+Better memory efficiency than v4
Scanpyv1.10+
clusterProfilerv4.10+
fgseav1.28+
STRINGdbv2.14+Uses STRING DB v12.0
Cytoscapev3.10+
GATKv4.5+
Cell Rangerv8.0+Official 10x Genomics release
Bioconductorv3.18+Assumes R ≥ 4.3

Each post states the versions current when it was written. When a tool gets a major update (which can break compatibility), we review the post and update it.

3. Code Verification Procedure

Every code example published on the blog goes through the following checks:

  1. Syntax check — confirmed to run without syntax errors in a real R/Python environment
  2. Intent check — confirmed on sample data that the code produces the intended result
  3. Version compatibility — confirmed to work on at least one recommended tool version
  4. Comment accuracy — every comment is checked against what the code actually does
  5. Dependencies stated — all required library() / import statements are included

Using the code examples

The code provided is for education and reference. When you apply it to real research, adjust the parameters to fit your own data and reproduce the results in your own environment. The code is free to use under the MIT License (no warranty).

4. Statistical Reporting Standards

Statistical analysis guides follow these reporting standards:

Differential Expression (DE) Analysis

  • State the multiple-testing correction method (Benjamini-Hochberg FDR by default)
  • Report both the p-value/FDR cutoff and the effect size (log2FC) cutoff
  • State the sample size (N per group)
  • State how batch effects were handled
  • State the pre-filtering criteria (e.g. removal of low-expression genes)

Functional Enrichment Analysis

  • Define the background (universe) clearly — whole genome vs. genes detected as expressed
  • Multiple-testing correction method (BH is standard)
  • Report both the FDR cutoff (usually 0.05) and the effect size (NES) cutoff
  • Version of the gene set database used (e.g. MSigDB v2023.2.Hs)
  • Leading-edge gene lists recommended as supplementary material

Survival Analysis

  • Report the results of testing the Cox proportional hazards assumption
  • Present the hazard ratio (HR), 95% confidence interval and p-value together
  • Validation in an independent cohort recommended (e.g. TCGA → METABRIC)
  • State the covariates in multivariable models

Biomarker Reporting

  • Follow the REMARK guidelines (Altman et al., 2012)
  • Follow the TRIPOD guidelines (prediction models, Collins et al., 2015)
  • Keep the discovery cohort and validation cohort clearly separate
  • Report clinical applicability together with limitations

5. Data Source Standards

Data sources used in the analysis guides:

CategoryRecommended sourcesLicense
Bulk RNA-seqTCGA, GEO, ArrayExpressPublic (CC-BY, etc.)
Single-cell RNA-seqSingle Cell Portal, Tabula Sapiens, HCAPublic or dbGaP
ProteomicsPRIDE, MassIVE, iProXPublic
Genomic variants1000 Genomes, gnomADPublic
ClinicalTCGA Pan-Cancer, MIMIC-III/IVAccess by application
Reference networksSTRING, BioGRID, Reactome, KEGGPublic (KEGG has commercial restrictions)
AnnotationEnsembl, GENCODE, RefSeq, UniProtPublic
Expression referenceGTEx, Human Protein Atlas, FANTOMPublic

Where possible, we give the GEO/SRA accession number or a download link for the datasets used in examples.

6. Code Sharing Policy

Code examples in blog posts are provided in the following forms:

  • Inline code blocks — copy directly from the text
  • Complete workflows — step-by-step code from input to results
  • Dependency information — library() / import statements included
  • Sample data — public data used where possible, or code to generate synthetic data included

We plan to collect larger workflows in a GitHub repository in the future. To request the full code for a specific post, get in touch via the Contact page.

7. Pitfalls Documentation

Every technical guide includes the following:

  • Common mistakes — traps beginners often fall into
  • Interpretation pitfalls — cases that are statistically significant but biologically meaningless
  • Tool limitations — which data and scenarios a tool does not handle
  • Alternatives — other methods to use when a tool hits its limits

This policy helps readers interpret analysis results critically.

8. Content Update Schedule

Recommended review intervals by content type:

  • Tool comparisons — reviewed every 6 months to 1 year (immediately when a tool has a major update)
  • Methodology guides — reviewed yearly (updated when best practice changes)
  • Benchmark results — updated when new benchmark papers are published
  • Database information — updated on major releases (e.g. STRING v12.0 → v12.5)
  • Clinical information — updated immediately when new guidelines are published

For significant updates, a "Last updated" date is shown at the top of the post.

9. Limitations and Disclaimer

Every analysis method has limitations. SBM Lab's guides present general best practice, but in a specific research context a different approach may be more appropriate.

Users are responsible for verifying the accuracy and applicability of any analysis results obtained using this blog's content. SBM Lab is not liable for any direct or indirect damages arising from results obtained through its content.

10. Policy Updates

This Methodology page is updated as the field develops. Major changes are noted at the top of the page, and a full review is carried out at least once a year.

Please send suggestions or questions about methodology via the Contact page.