Neoantigen × AI
Daily research signal
№60Brief

Personalized melanoma vaccine cuts recurrence, death risk 42%

Personalized neoantigen vaccines hit two milestones today: a phase 3 efficacy readout combined with Keytruda in high-risk melanoma, and phase 1 evidence that individualized vaccines can be manufactured and delivered in ovarian cancer.Anadolu AjansıMedical Xpress

Underneath both sits the prediction layer, where a new benchmark shows that decoy construction — not just model quality — shapes reported pMHC predictor performance, a caution for anyone building or diligence-ing neoantigen selection pipelines.bioRxiv

3new stories
ClinicalyesterdayNew

Personalized ovarian cancer vaccines spark immunity in phase 1 trial - Medical Xpress

A phase 1 trial of NeoVax in high-grade epithelial ovarian cancer showed that personalized neoantigen vaccines could be successfully manufactured and administered alongside standard chemotherapy and immunotherapy, in both newly diagnosed and recurrent patients, with the reported signal being induced immunity. Each vaccine was custom-built as a pool of neoantigen-targeting peptides identified by sequencing DNA and RNA from a patient's tumor and healthy cells. The result extends the personalized neoantigen approach into ovarian cancer, though at phase 1 the data speak to feasibility and immunogenicity, not efficacy.

news · yesterday · Medical Xpress

02

01
AI / Methods2 days agoNew

BenchMHC: a reproducible framework for peptide-MHC presentation prediction

BenchMHC, an open framework for reproducing, retraining and evaluating peptide-MHC presentation models on public data, matched the official NetMHCpan-4.1 webserver (0.794 vs 0.790 per-allele Mean Top-K; 0.0025 vs 0.0022 Median FRANK) while scoring 280,000 peptide/allele pairs in 27 seconds, roughly 350x faster than the original binary. Its more pointed finding: 86-99% of hits in the public eluted-ligand training sets appear in the human proteome versus about 12% of decoys, so models can separate hits from decoys on proteome frequency rather than MHC-binding features. Redrawing decoys within each hit's source protein and matching cysteine status, with no other change, raised per-allele Mean Top-K on the regenerated evaluation. For neoantigen pipeline builders, the takeaway is that benchmark and decoy design can confound method comparisons and inflate apparent predictor gains.

biorxiv · 2 days ago · Borck, T.; Chabrol, M.; Colombo, D.; Imbert, A.; Perier-Dulhoste, M.; Volzhenin, K.; Lang,
02
Clinical6 days agoNew

Personalized cancer vaccine cuts risk of melanoma recurrence or death by 42% in trial - Anadolu Ajansı

A phase 3 trial in more than 1,000 patients with high-risk melanoma reported that a personalized cancer vaccine added to Keytruda cut the risk of recurrence or death by 42% versus Keytruda plus placebo, and reduced the risk of melanoma spreading by 38% over two years. Treatment ran roughly one year in a double-blind, international design, positioning the vaccine as an adjuvant therapy aimed at tumor cells remaining after surgery. The report does not name the vaccine's developer, but a positive phase 3 readout in this setting is a significant step for the personalized neoantigen approach.

news · 6 days ago · Anadolu Ajansı