Epitope Prediction
The Bulk Epitope Prediction pipeline submits sequences to the FIRE API for batch epitope binding prediction. It follows the same polling pattern as Bulk Similarity Search but returns predicted epitope scores instead of nearest-neighbor matches.
When to use
Section titled “When to use”Use this pipeline when you want to predict epitope specificity for a set of TCR or BCR sequences. FIRE’s epitope prediction model scores each sequence against a panel of known epitopes.
Web App: Submitting a job
Section titled “Web App: Submitting a job”Navigate to Dashboard → FIRE Bulk Epitope → New Job (or via Experimental → Single Epitope Prediction for individual sequences).
Required inputs
Section titled “Required inputs”| Field | Description |
|---|---|
| Job name | A label for this run |
| Input CSV files | One or more CSV files with a sequence column, uploaded as a zip archive |
| Sequence column | Column name containing sequences to predict |
| Epitope panel | Name of the FIRE epitope panel to predict against |
Optional inputs
Section titled “Optional inputs”| Field | Default | Description |
|---|---|---|
| Score threshold | none | Minimum prediction score to include in results |
| Include all epitopes | false | If true, returns scores for all epitopes in the panel (not just top matches) |
Output files
Section titled “Output files”results.csv
Section titled “results.csv”Input CSV with appended epitope score columns:
| Column | Description |
|---|---|
epitope_<name>_score | Predicted binding score for epitope <name> (0–1) |
top_epitope | Epitope with the highest predicted score |
top_score | Highest predicted score |
failed.csv
Section titled “failed.csv”| Column | Description |
|---|---|
sequence | Original query sequence |
error | Error description |
Experimental: Single Sequence Epitope Prediction
Section titled “Experimental: Single Sequence Epitope Prediction”For one-off predictions without submitting a full pipeline job, use the Experimental → Single Epitope Prediction page in the web app. Enter a single sequence and view instant results from the FIRE API.
See Experimental Features for more.
Troubleshooting
Section titled “Troubleshooting”Unknown epitope panel name
Panel names are case-sensitive. Contact the FIRE API administrator for the list of available panels, or check the FIRE docs.
All scores are near zero
The sequence may not match the receptor type expected by the panel (e.g., submitting BCR sequences to a TCR panel). Verify sequence type compatibility with the selected panel.