The count of predicted or experimentally presented neoantigens in a tumor — a more biology-aware cousin of tumor mutational burden that estimates how many mutations actually yield immune-visible targets.
Tumor mutational burden (TMB) counts all somatic mutations per megabase as a proxy for how many neoantigens a tumor might generate. But most mutations never become visible to the immune system: only a fraction fall in coding regions, alter the protein, produce a peptide that is processed and HLA-presented, and is recognized as foreign. Tumor neoantigen burden (TNB) tries to count the subset that survives this filter — typically the number of predicted neoantigens (often HLA-presentation-filtered) or, more rigorously, peptides confirmed by immunopeptidomics.
Because TNB targets the immunologically relevant quantity directly, studies have reported it as a better correlate of immunotherapy response and prognosis than raw TMB in several cancers. The catch is dependency: TNB is only as good as the prediction stack that produces it — HLA typing, binding/presentation prediction, and (ideally) immunogenicity filtering all feed in, so errors compound and there is no universally agreed pipeline or cutoff. TMB, by contrast, is cheaper and more reproducible because counting mutations needs no immunology model.
So what for investors: TNB is a more mechanistically honest biomarker than TMB and a natural selling point for companies whose edge is better presentation/immunogenicity prediction — their model directly defines the biomarker. But it is also less standardized and harder to validate clinically. When evaluating a TNB-based claim, scrutinize the underlying pipeline (which predictor, EL vs BA, immunogenicity filtering, HLA coverage) and whether the cutoff was prospectively defined, not fit after the fact.
TMB counts all somatic mutations per megabase; TNB counts only the mutations predicted (or shown) to yield presented neoantigens. TMB is a cheap, reproducible proxy for potential immunogenicity, while TNB attempts to measure the immune-visible subset directly — making it more biologically meaningful but dependent on prediction accuracy.
Several studies report TNB correlating better with response and prognosis than raw TMB, because it targets the immunologically relevant antigens rather than total mutations. However, TNB is not yet standardized — results depend heavily on the prediction pipeline and chosen thresholds — so it has not displaced TMB as a routine clinical biomarker.
Typically by calling somatic mutations, translating mutated peptides, HLA-typing the patient, and running HLA-presentation prediction to count likely neoantigens; some pipelines add immunogenicity filtering or confirm peptides directly via mass-spec immunopeptidomics. Because each step introduces error, TNB values vary with the tools and cutoffs used.