NEW YORK (GenomeWeb) – Based on information in tens of thousands of tumor genomes, a research team from Foundation Medicine and Roche Pharma Research & Early Development concluded that it would be unlikely to come up with a cancer vaccine that would benefit more than a fraction of the population.
Teams have started delving into the mutations present in an individual's tumor genome in an effort to come up with personalized cancer vaccines. But despite the mutational diversity that's been described in large-scale sequencing efforts, there is also interest in developing vaccines that might be more broadly applicable.
For the new study, the researchers searched thousands of tumor samples for common, neoantigen-producing mutations that might be targetable with a semi-universal cancer vaccine. Based on the alterations identified — with the help of information on two common human leukocyte antigen (HLA) subtypes — they estimated that such a vaccine would benefit only about 0.3 percent of the population.
Although some genetic changes did turn up relatively often, computer modeling indicated that a relatively small proportion are predicted to produce neoantigens that would be recognized as non-self molecules by the immune system, Foundation Medicine researcher Ryan Hartmaier said in a statement.
Hartmaier, who was first author on the team's Genome Medicine paper published online today, said "we were able to select a panel of 10 neo-antigens that could be applied to the maximum amount of unique tumors in our data set." But he noted that just 0.7 percent to 2.5 percent of the tumors considered appeared to contain one or more predicted neoantigens. And that applicability dipped further in the broader population when available HLA frequency data was taken into account.
The researchers assessed coding sequences from hundreds of cancer related genes, nabbed by hybrid capture from 63,220 solid tumors or hematologic cancers, and sequenced to more than 500-fold average coverage with Illumina instruments.
The team focused on recurrent mutations or recurrently mutated genes, using an in silico prediction approach to search for common alterations that produce peptides expected to bind to class I or II major histocompatibility complex molecules in the immune system — predictions predicated on the most common HLA haplotypes in North American or European populations.
The researchers whittled more than 2,800 detected recurrent mutations down to sets of predicted neoantigens recognized by one or both of the two main HLA subtypes. Even so, they estimated that even in the 'best-case scenario,' those neoantigens would not be broadly applicable.
They noted that the collection of neoantigens considered in future studies may be influenced by addition factors such as the actual transcription and translation patterns of mutated genes, tumor clonality and selection, HLA frequencies in a given population, and so on. Moreover, the authors explained that the set of suspected neoantigens identified may differ in tumor sequences analyzed alongside those from matched normal samples as well as tumors subjected to more extensive exome sequencing or analyses.
"It is possible to identify a set of alterations shared across patient tumors for the production of non-individualized, poly-neoantigen cancer vaccine in an HLA subtype-specific manner," they wrote. "However, with current neoantigen prediction methodologies, this approach will be applicable to only a small proportion of the population."