Every candidate is scored, weighed, and most are set aside. Only the strongest evidence earns a place in the vaccine.
The Neoantigen Selection Platform is the decision-making core between everything a tumor's sequencing turns up and what the company actually manufactures.
Whole exome and RNA sequencing of the tumor and matched normal tissue.
Identifying what has genuinely changed in the tumor's DNA relative to the patient's own healthy tissue.
Determining which peptide shapes this patient's immune system is actually able to present.
Confirming a mutation is actually expressed by the tumor, not just present in the DNA and silent.
Binding strength, immune visibility, antigen-presentation machinery integrity, stemness weighting and manufacturability, each scored independently rather than blended into one number.
A final, capped set of peptides for manufacturing, with an explicit, logged reason for every candidate that did not make the cut.
A qualified reviewer approves or overrides the ranked list before anything moves toward manufacturing.
Most neoantigen-selection pipelines score candidates almost entirely on binding affinity and gene expression: how strongly a peptide sticks, and how much of it is present.
NSP weights candidates by whether the originating mutation is present in the tumor's stem-like, treatment-resistant cell population, not just the bulk tumor.
Bulk tumor expression can be dominated by cells that are already dying off or being cleared by treatment. The cells that actually cause relapse are the stem-like ones. A vaccine that targets mutations specific to that population is a bet on preventing recurrence, not only shrinking what is visible on a scan today.
Every signal in NSP is its own swappable, self-contained term, not baked into one formula.
A newly published predictor or marker can run as a fully audited "shadow" configuration on real cases in parallel, never touching a real patient's vaccine until it is validated and deliberately promoted.
New neoantigen sources plug into the same scoring core untouched.
Today's published neoantigen-prediction models are trained on data skewed toward alleles common in European-ancestry populations, a gap that matters for Canada's genetically diverse patient population. NSP is explicitly tested and improved across a broader range of HLA types.
The asset is the trained, validated selection model and its outcome dataset, not the licensed vaccine chemistry itself. The moat is the stemness-weighted scoring layer and the modular, shadow-tested architecture, not any single formula. A future direction for the company's intellectual property is pairing predicted neoantigens with adjuvant selection: predicting which peptide and adjuvant combinations perform best together.
See the field's evidence base