Neoantigen discovery: Overcoming the challenges of neoantigen prediction
This is Part II of a two-part series on the development of the Personalis Neoantigen Discovery Report. Part I, Prediction Challenges can be found here.
At Personalis, we recognize the challenges of neoantigen prediction and are working to address each of them. Our current pipeline (see Figure 1) is designed to accurately rank neoantigens using DNA and RNA sequencing data derived from the analytically-validated, augmented ACE Exome and ACE Transcriptome. The pipeline focuses on accurately detecting neoantigens based on peptide processing, MHC binding prediction, similarity to self, similarity to known antigens, and immunogenicity. Our reports provide a cohesive breakdown of the somatic mutations, resulting peptides, and each of these important considerations.
Figure 1. The diagram above shows our neoantigen identification process, which is downstream of our DNA and RNA cancer pipelines.
To evaluate pipeline performance, the team began with a proof-of-principle experiment using well-established peptides known to be immunogenic and elicit a T-cell response (taken from Cancer Immunity). Briefly, these peptides were reverse engineered to generate the derived variants and in silico spiked into a well characterized cancer cell line. After analysis, 22 out of the 23 peptides were accurately detected by our pipeline, resulting in an overall sensitivity at detecting known immunogenic peptides of 96%.
This characterization is just the beginning. We are currently working hard to implement solutions for each of the aforementioned immuno-oncology challenges, including performing Class II MHC binding predictions, the importance of detecting other alterations such as frameshift InDel-derived or out-of-frame fusion-derived peptides, phasing of neoantigens, tumor escape mechanisms, and assessing the tumor microenvironment. Many of these important features are nearing integration into our pipeline.
Moving forward we aim to continually improve not only our pipeline but also, very importantly, our validation. We will continue to leverage previously identified neoantigens, increasing our number and breadth while expanding into new somatic mutation classes. Additionally, we will also be applying cutting-edge proteomic and immunogenic approaches to assess MHC binding and immunogenicity on real patient samples, using the results to continually improve our pipeline.
Our objective is to create a comprehensive neoantigen bioinformatics solution to enable accurate assessment of immunotherapies with likelihood of durable clinical responses. We envision our solution being used for the development of personalized vaccines, as well as retrospective analysis of vaccine clinical trials, to better understand what differentiates patient responders from non-responders, which is currently a high priority unmet need. Additionally, as translational research has become an increasingly crucial part of the development process, we hope that by providing a broad view of the I-O landscape through our report and assay design, we can play a role in understanding combination trials (i.e. vaccine + checkpoint modulator) to assist in identification of effective combinations. Determining who will respond to treatment and what will be the most effective therapeutic combination for them would provide insight to treatment selection in a sea of potential PD-1 and PD-L1 combination therapies, and therefore both pharma companies and patients have much to gain.