Most previous vaccination attempts in cancer immunotherapy have been focused on a single antigen. Consequently, the expression of the selected antigen needs to be determined before trial inclusion resulting in a high screening failure rate. The establishment of a multi-antigen vaccine may improve the coverage of potential patients, especially if the selected antigens are semipersonalized for clinical patient subgroups.
Analysis and Visualization of Alteration Data (AVATAR), a novel bioinformatic software tool for multi-objective optimization of large datasets, was used to analyze publicly available datasets of transcriptome data from The Cancer Genome Atlas (TCGA) and a dataset from Gene Expression Omnibus (GSE 40774). Cancer-testis antigens (CTA) were selected as model antigens for multi-objective optimization based on the primary tumor site and HPV-status. Dichotomized data were pooled and edited resulting in a cohort of 865 patients, of which 125 were HPV-positive (HPV+). Primary tumor sites included were oral cavity (n = 208), oropharynx (n = 150) and larynx (n = 208).
Selecting 10 CTA genes for the whole cohort results in a coverage of 89.3%. When optimizing gene selections for each primary tumor site, distinct 10-gene-selections are revealed improving the coverage in the respective group markedly. Antigen selections overlap only to a minor degree with the solutions for all patients or other patient groups. Even in predominantly HPV- primary sites such as OC and L, antigen selections are remarkably different from each other. When optimizing for HPV-status, distinct selections were found for optimal coverage in the respective patient group overlapping only in one gene.
AVATAR can be used to identify antigen selections with optimal coverage within a specified patient group. A semi-personalized antigen selection based on multi-objective optimization may be a useful strategy to plan trials for antigen-specific vaccination.
Clinical trial identification
Legal entity responsible for the study
University Medical Center Ulm.
Deutsche Forschungsgemeinschaft (GRK2254).
S. Laban: Advisory board member: AstraZeneca, Merck Sharp & Dohme; Lecture fees: Bristol-Myers Squibb, Merck Serono. G. Völkel, H. Kestler, J. Kraus: Developers of Analysis and Visualization of Alteration Data (AVATAR) software. T.K. Hoffmann: Advisory board member: MSD, Merck Serono. H.-G. Rammensee: Cofounder and shareholder: Immatics, CureVac, Synimmune. P.J. Schuler: Advisory board member: Bristol-Myers Squibb. All other authors have declared no conflicts of interest.