Prediction and prioritization of neoantigens: integration of RNA-Seq data with whole-exome sequencing

Abstract

The importance of neoantigens for cancer immunity is now well-acknowledged. However, there are diverse strategies for predicting and prioritizing candidate neoantigens, and thus neoantigen loads reported in the literature vary a great deal. To clarify this issue, we compared the numbers of neoantigen candidates predicted by four currently utilized strategies. Whole-exome sequencing (WES) and RNA-Seq of 4 non-small cell lung cancer patients was performed. We identified 361 somatic missense mutations from which 224 candidate neoantigens were predicted using MHC class I binding affinity prediction software (Strategy I). Of these, 207 exceeded the set threshold of gene expression (fragments per kilobase of transcript per million fragments mapped (FPKM) ≥1) resulting in 124 candidate neoantigens (Strategy II). To verify mutant mRNA expression, sequencing of amplicons from tumor cDNA including each mutation was performed; 204 of the 207 mutations were successfully sequenced, yielding 121 mutant mRNA sequences, resulting in 75 candidate neoantigens (Strategy III). Sequence information was extracted from RNA-Seq to confirm the presence of mutated mRNA. Variant allele frequencies ≥0.04 in RNA-Seq were found for 117 of the 207 mutations and regarded as expressed in the tumor, and finally, 72 candidate neoantigens were predicted (Strategy IV). Without additional amplicon sequencing of cDNA, Strategy IV was comparable to Strategy III. We therefore propose Strategy IV as a practical and appropriate strategy to predict candidate neoantigens fully utilizing currently available information. At any rate, it is of note that different neoantigen loads were deduced from the same tumors depending on the strategies applied.

This article is protected by copyright. All rights reserved.

http://ift.tt/2hKQ0rJ