Human cancer immunotherapy research
Explore an analysis.
Follow the evidence from your inputs to a candidate shortlist.
Public viral peptide examples exercise binding prediction and ranking. They are not patient mutations or personalized cancer vaccine targets; no expression values or clinical outcomes are supplied.
IEDB epitope search ↗Start with peptide and HLA inputs.
Each row pairs a protein fragment with an exact HLA allele. Supply expression only when you have comparable measurements.
Tab-separated columns: peptide, allele, optional expression, optional source ID. Maximum 200 pairs; canonical 8–14 amino-acid peptides.
- 1
Sequencing & variant calling
Find genetic changes in tumor DNA.
Upstream of this peptide input - 2
Variant annotation
Understand which genes and proteins may be affected.
Upstream of this peptide input - 3
Mutant peptide construction
Generate fragments the immune system might recognize.
Upstream of this peptide input - 4
MHC binding prediction
Estimate binding to the supplied HLA molecules.
MHCflurry · Class I affinity - 5
Neoantigen prioritization
Compare evidence and build a shortlist.
Transparent evidence-based ranking - 6
Experimental validation
Connect selected candidates to laboratory research.
Export for experimental handoff
Review your analysis
Run fresh binding predictions and rank the supplied candidates.
- Input rows
- 3
- Entry point
- MHC binding prediction
- Model
- MHCflurry class I affinity
Methods & settings
Lower predicted affinity in nM indicates stronger predicted binding. Ranking uses within-run binding percentiles; it is not a probability of immune recognition.
At zero, rank by binding alone. Above zero, rows missing expression are excluded from the combined policy.
Up to 200 peptide–HLA pairs per analysis. Every completed run saves its inputs, settings and results.
Your results will live here.
Start with your inputs, then run an analysis. Every completed run keeps its own evidence and settings.