FAQ
Is this a clinical diagnostic tool?
No. Oncology in Gradient Biotech is a research and translational analysis workspace. Outputs must not be used for clinical diagnosis, treatment selection, or regulatory decision-making. This is not a replacement for expert oncology review.
Do I need R, CellPhoneDB, or maftools installed locally?
No. Upload files through the web UI and the platform handles analysis. Reference implementations inform the underlying algorithms but you never need to run them yourself.
How do I run my first pipeline?
Create a study, upload a MAF or clinical CSV (see Sample data), and run the matching pipeline from its workflow page. See Quick start.
Cell communication failed — what should I check?
Make sure both the expression and metadata files are selected and that they share matching cell_id values, with a cell_type column in the metadata file.
Mutation landscape failed — what should I check?
Confirm your mutation file is selected and is a valid MAF-style CSV with the required columns (Hugo_Symbol, Tumor_Sample_Barcode, Variant_Classification, and position/allele columns).
Why are some mutations "not placed" in the 3D view?
Each position is checked against the structure: the reference amino acid in the protein change must match the residue there. Mismatches usually mean the upload uses a different protein isoform than UniProt canonical; positions beyond the modelled sequence or without a residue number are also listed. Nothing is placed on a residue that does not match. See 3D protein structure.
Why is "Find 3D hotspots" disabled or showing no statistics?
The hotspot analysis runs on the AlphaFold model, so it is unavailable for proteins AlphaFold DB does not cover (for example BRCA2 and NF1, which are viewed on experimental structures instead). Statistics are also withheld when fewer than 5 confidently modelled residues are mutated, or when more than 10% of positions fail validation.
Do structures come from third-party services?
Yes — AlphaFold DB, the Protein Data Bank (PDBe/RCSB), and UniProt — but they are fetched and cached by the platform's server, never from your browser, so your gene queries are not sent to those services.
Survival analysis failed — what should I check?
Confirm your clinical file is selected and has patient_id, time_to_event_days, and event_observed columns.
Interpretation failed — source runs not found
The runs you selected for interpretation must still be complete and available. If one was deleted after you started, re-run the interpretation with currently available runs.
How does oncology relate to Computational Biology?
Computational Biology owns single-cell clustering, spatial transcriptomics, differential expression, and biomarker ML. Oncology adds cell-cell communication, immune deconvolution, mutation landscape, survival analysis, and clinical endpoint integration. TME profiling calls compbio infrastructure rather than duplicating it.
How does oncology relate to Pathology?
Pathology owns whole-slide imaging, tissue segmentation, and spatial quantification. Oncology uses pathology outputs for pathology-integrated spatial TME analysis without rebuilding the image processing layer.