Abstract 96P
Background
Non-hotspot mutations in oncogenes are frequently identified during complex molecular profiling. Lack of functional or clinical data complicates interpretation of such mutations as driver or passenger event. Computational algorithms may be used to define the role of previously not annotated mutation in cancerogenesis and therapy decision.
Methods
We used JAX and OncoKB databases to collect annotations of known mutations. Computational predictions were performed employing numerous methods, including SIFT, ProVean, CADD, VEST4, CHASM, FATHMM, REVEL,MutPred, MetaLR.
Results
We collected a total of 906 mutations with consistent annotation across JAX and OncoKB, including 763 oncogenic and 203 neutral mutations in selected 39 oncogenes. We defined three sets of stringent criteria based on prediction results of diverse methods which allowed high-confidence prediction of neutral status of mutation. These allowed correct prediction of neutral status of 41 (23%) neutral mutation while erroneously predicted as neutral 5 (0.6%) mutations annotated as oncogenic in JAX/OncoKB, including EGFR p.I491M, p.G465R, p.S492R. Latter are known to prevent binding of cetuximab resulting in therapy resistance, while no data about oncogenic status of these variants were published. Additionally we defined two sets of criteria for highly confident prediction of oncogenic status which allowed correct prediction of oncogenic status of 106 (14%) oncogenic mutations with single (0.5%) neutral mutation erroneously predicted as oncogenic. We applied defined sets or rules to 130 unique mutations, identified across 554 patients referred for complex molecular profiling. This allowed to get high-confidence prediction of neutral status of 20 mutations (100% are non-hotspot mutations) and oncogenic status of 44 mutations, including 5 non-hotspot mutations (including ALK p.S1487L, RET p.W557I, KIT p.S746L, ERBB4 p.L428H and MET p.S637F mutations).
Conclusions
In routine practice of complex molecular profiling computational methods allow high-confidence prediction of oncogenic/neutral status for 30% of non-hotspot mutations.
Clinical trial identification
Editorial acknowledgement
Legal entity responsible for the study
The authors.
Funding
Has not received any funding.
Disclosure
V. Yakushina, A. Lebedeva, G. Timokhin, E. Veselovsky, V. Mileyko, M. Ivanov: Financial Interests, Personal, Full or part-time Employment: Atlas oncologyDiagnostics. All other authors have declared no conflicts of interest.