11589 Background: Neoantigen identification is increasingly critical for clinical immuno-oncology applications including predicting immunotherapy response and neoantigen-based personalized cancer vaccines. Although standard research pipelines have been developed to aid neoantigen identification, building a robust, validated neoantigen identification platform suitable for clinical applications has been challenging due to the complex processes involved. Methods: To improve neoantigen identification, we extended standard sequencing and informatics methods. We developed an augmented and content enhanced (ACE) exome sequenced at 200X to increase sensitivity to SNPs and indels used for neoantigen identification as well as HLA performance. To accurately identify fusions and variants from RNA, we optimized our ACE transcriptome for FFPE tissue. To improve neoantigen pipelines based on MHC binding algorithms, we developed peptide phasing, high accuracy HLA typing, TCR interaction predictors, and transcript isoform estimation tools to detect neoantigens from indel and fusion events. We performed comprehensive analytical validation of the platform including the ACE Exome, somatic SNV/indel calls, RNA based variant and fusion calls, and HLA typing. This was followed by an overall in silico validation of neoantigen identification using 23 experimentally validated immunogenic neoepitopes spiked into exome data. Results: Analytical validation of our ACE exome platform showed > 97% sensitivity for small variants with a specificity of > 98% at minor allele frequency > 10%. From the ACE transcriptome we achieved a fusion sensitivity of > 99% and RNA based variant calls sensitivity of > 97%. Our ACE exome based HLA typing was 98% and 95% concordant with Class I and II HLA results (respectively) from clinical testing. Our in silico validation of neoantigen predictions resulted in identification of 22 out of 23 immunogenic neoepitopes. Conclusions: We developed sequencing and informatics improvements to standard approaches that can enhance neoantigen identification and demonstrated a comprehensive validation approach that may support neoantigen use in future clinical settings.
Requirements of integrated computational approach for developing personalized cancer vaccines
