For all the promises of a fresh new approach to cancer vaccines and therapy – zeroing in on specific mutated antigens expressed only by tumors – companies and academics have little way of assessing how well they are able to predict. Represent the best goals. There is no standard or baseline for players to stack themselves against rivals in the nascent zone, and by the time they find out, it may be too late.
No group could really build that benchmark. Precious proprietary information is at stake, not to mention the necessary resources.
But four years ago, as the concept was coming to an end, it attacked the Parker Institute for Cancer Immunotherapy as its collaborative model was created to solve the problem.
“We are like Switzerland of neoantigens,” Danny Wells, chief data scientist at PICI told Endpoint News.
So working with the Cancer Research Institute and a non-profit organization called Sage Bayon Networks, it brought together more than 40 biopharma companies and educational laboratories, gave them the same melanoma and non-small cell lung cancer tissue, and Asking each team to present its most promising neonatigen predictions. . The PICI researchers then went to the lab and compared the predictions, testing whether neoantigens were actually recognized by T cells. The result is a basic dataset, named TESLA, which is an acronym for the Tumor Neoantigen Selection Alliance – making it public to the scientific world today.
To the researchers’ surprise, the differences between prediction algorithms were “tremendous”, said Wells, a co-senior author in the paper published in Cell. No team managed to identify every renaissance or a large majority of them: “The overlap between predictions, no matter how we did it, was actually low.”
He said the need for new knowledge in the field has been highlighted, and PICI believes that TESLA has gained some insight.
A set of five distinct features, it turned out, can predict good neonatings with high accuracy and specificity when integrated into a model. They are ligation affinity, tumor abundance, and ligation stability, which has to do with how neoetigens are presented; As well as agretopicity and foreignness, which is related to their recognition by immune cells.
“These are all the features that were talked about, but I think we were surprised that by simply integrating them together in this single model that it works so well,” Wells said.
When participating teams re-applied these features to their algorithms, PICI reported, predictions improved. A data model emphasizing all five features derives from a test against another set of cancer samples, with 75% effective newborn targets being accurately estimated and 98% ineffective people filtered out. Wells said drug developers may test their algorithms differently depending on the therapeutic strategy. But the hope is that he and co-senior writer Nadine DeFranoux have it that they can provide a common baseline for both those who are already in the field and others who want to jump into it.
“This research has the potential to improve the mathematical algorithms of drug manufacturers and researchers,” said Lisa Butterfield, vice president of research and development at PICI. “This may prioritize the possibility of antigens being present on each patient’s cancer and are most visible to the immune system that are not.”
Although TESLA will not monitor the space for long to see how the companies compare over time, Wells envisions new proposals for the alliance. Many questions still need to be answered; For example, when they focused on class 1 prediction, or how CD8 + T cells view tumors, research by co-senior author Robert Schreiber suggested that class 2, or CD4 + T cells, would be a tumor. How she sees it is equally important.
“So far, neonatal predation has been a black box,” said Schreber, a professor at Washington University School of Medicine in St. Louis. TESLA has begun to shed light on this, and intends to continue it.