Gritstone presents data on indentifying MHC class 2 antigens at AACR

Gritstone Oncology announced the presentation of data demonstrating its ability to identify major histocompatibility complex, or MHC, class 2 antigens at the American Association for Cancer Research, or AACR. In an oral presentation, it was shown that Gritstone Edge improved the positive predictive value for human leukocyte antigen class II, or HLA-DR, peptide presentation over standard methods by approximately 20-fold. Edge is an artificial intelligence platform that identifies tumor-specific neoantigens, or TSNA, for the development of antigen-directed immunotherapies that may drive highly specific tumor cell destruction by T cells. TSNA can be presented by either MHC class I, which are recognized by CD8 T cells, or MHC class II, which are recognized by CD4 T cells. The public tools available to predict tumor-specific antigens presented by MHC class I are more advanced; historically, the characterization of MHC class II presented antigens has been challenging for the field due to greater variability in their binding properties. The MHC class II dataset for the AACR analyses was derived from 73 human tumor and cell-line samples, including non-small cell lung cancer, lymphoma and ovarian cancer. The data sat is comprised over forty-five thousand tumor-presented peptides. Building on the progress with class I Edge, Gritstone’s class II model overcame a challenge with HLA class II prediction, which is the longer and more variable presented peptide lengths. Gritstone addressed this challenge with the new comprehensive training dataset and an innovative neural network architecture, leading to an approximately 20-fold increase in performance when using Edge versus standard methods. This dataset complements the previously reported Edge data demonstrating that it is approximately nine-fold better than publicly available tools at predicting tumor-specific antigens presented by MHC class I.
https://thefly.com/landingPageNews.php?id=2887651