Computational genomics & reproducible analysis · 2026
Independent reproduction of cancer genomics studies
Rebuilding key analyses from recent synthetic lethality and renal cancer evolution studies using their source data and author workflows.
Two published studies rebuilt from their source data and author workflows
Approach
How far can key findings from recent cancer genomics studies be traced and reproduced from the available data and code?
Reproducing a genomics study requires more than redrawing its figures. This work follows each analysis from its source data and code to the published result, separating direct replots, statistical refits and reruns of author workflows. It also records where processed inputs, software or historical data limit exact reproduction.
- Trace data and author code
- Rerun and refit analyses
- Audit outputs and differences
Synthetic lethality and tumour expression
Original study: Haider et al., Nature Genetics · 2025
SYLVER examines whether expression changes in tumours reflect synthetic lethal relationships linked to tumour-suppressor loss. I built an executable Python workflow covering the main quantitative panels from Figures 2 to 4, with source-data links and interpretation for each panel.
The reproduction used the published source data, refitted patient-level clinical models and reran the supplied RB1 and BRCA analyses on processed workspaces. Differences and modelling choices were recorded in a separate register.
- RB1 intersection genes recovered
- 74
- BRCA1/2 intersection genes recovered
- 18
- HRD signature genes recovered
- 73
| Work completed | |
|---|---|
| Published source data | Reproduced main quantitative figure panels |
| Clinical analyses | Refitted patient-level models |
| Processed workspaces | Reran supplied RB1 DepMap and TCGA BRCA analyses |
| Quality control | Tracked sources, validation checks and analysis differences |
Non-genetic evolution in clear cell renal cancer
Original study: Fernández-Sanromán et al., Cancer Discovery · 2025
TRACERx Renal studies non-genetic evolution from primary to metastatic clear cell renal cell carcinoma. I organised an R workflow around the order of Figures 1 to 6, using retained numerical inputs and author code to rebuild data-backed views.
Each view was linked to its inputs and export records, with panels that could not yet be recreated clearly marked. Author-code replays were kept separate from article-aligned sensitivity analyses.
- Data-backed views rebuilt
- 34
- Main figures followed
- 6
| Work completed | |
|---|---|
| Figure workflow | Rebuilt available panels in the published figure order |
| Author workflows | Reran selected RNA expression, immune-state and tumour-evolution analyses |
| Traceability | Recorded plotted data, source hashes and figure settings |
My contribution
I organised the source evidence, built the Python and R workflows, reran and refitted selected analyses, checked outputs against published figures and tables, and documented differences and sensitivity analyses.
Study scope
This is independent reproduction work and not an original or published study. The linked papers belong to their respective authors. Results cover selected analyses and depend on the supplied data and code, so they do not replace the original studies or provide new clinical validation.