Synthetic lethality–informed neoantigen prioritisation to support mechanism-specific therapeutic strategies across human cancers.

2510 Background: Personalised neoantigen vaccines can elicit strong tumour-specific T cell responses, yet tumours may evade immune pressure by deleting or silencing the targeted gene. Current selection pipelines treat gene-based escape as an unpredictable failure. For tumour suppressor genes, however, loss of the target can create predictable synthetic lethal (SL) dependencies. We present a framework that classifies tumour suppressor mutations by their therapeutic implications. We hypothesised that hypomorphic mutations, which generate immunogenic neoantigens while retaining partial function, may be suited to sequential therapy, where vaccination selects for complete loss and exposes druggable vulnerabilities. In contrast, null mutations with pre-existing SL dependencies may benefit from upfront vaccine–drug combinations. Methods: We analysed 9,953 tumours across 32 cancer types from TCGA. Tumour suppressor mutations were assessed for neoantigen potential using NetMHCpan-4.2, MHCflurry 2.0, and PRIME 2.0 (%rank < 2%), with filtering for clonality, expression (TPM > 1), and multi-allele HLA presentation. Functional status was assigned using gene-specific annotations, including TP53 transactivation scores. SL relationships were curated from DepMap, SynLethDB 2.0, and published clinical evidence, restricted to partners with approved or Phase II+ inhibitors. Results: 7.4% of patients (n = 741) harboured tumour suppressor neoantigens with matched SL vulnerabilities, which segregated into two distinct groups. The first group (4.6%, n = 458) carried hypomorphic mutations that generated immunogenic neoantigens while retaining partial gene activity, consistent with candidacy for sequential vaccination followed by SL inhibition upon escape. TP53 missense mutations predominated (58%), with progression to null status predicted to confer WEE1 or CHK1 vulnerability. Additional recurrent pairings included PTEN–AKT (14%), ATM–ATR or PARP (11%), KMT2D–EZH2 (9%), and BRCA2–PARP (8%). The second group (2.8%, n = 283) carried functionally null mutations with SL dependencies present at diagnosis, nominating these patients for upfront vaccine–drug combinations. Endometrial (12.1%) and colorectal (9.8%) cancers were enriched for both patterns, whereas high-grade serous ovarian cancer was dominated by hypomorphic mutations. Conclusions: This framework reframes immune escape from neoantigen vaccination, conventionally viewed as treatment failure, as a predictable and potentially exploitable therapeutic event. Functional classification of tumour suppressor neoantigens provides a rationale for distinguishing sequential from upfront vaccine–SL inhibitor strategies. Prospective trials with longitudinal monitoring will be required to determine whether vaccination-induced gene loss consistently exposes the predicted vulnerabilities.