02/24/2021 | News release | Distributed by Public on 02/24/2021 14:39
We are witnessing one of the most accelerated scientific efforts in response to the COVID-19 pandemic. Among these efforts that include sequencing the virus, designing the very first prophylactic vaccines, conducting clinical trials, demonstrating efficacy, and getting regulatory approvals in record time, many companies have leveraged their vaccine technology and infrastructure to help combat the current pandemic. While a 30,000 bp virus is relatively simple in nature to understand in terms of structure, function, and immunogenicity, applying the same technology and lessons learned to design vaccine therapeutics for complex diseases like cancer is a much more demanding task.
Indeed, the oncology field has been investigating cancer vaccines for years, investing specifically in neoantigen-based designs targeting solid tumors. The concept of packaging immunogens from the patient tumor into a vaccine to mount an immune response against the tumor has materialized in the form of numerous designs and approaches currently in clinical trials. To name a few examples below:
Challenges of Neoantigen identification and selection
Somatic mutations are a hallmark of cancer and result in mutant peptides, that are called neoantigens. These neoantigens are immunogens considered tumor-specific and foreign such that they can be recognized and efficiently eliminated by the host's immune system. Therefore, neoantigens became ideal targets for therapy design including cancer vaccines and cell therapies. The first step in building an efficacious neoantigen-based cancer vaccine is to accurately identify, select, and prioritize immunogenic neoantigens. This is achieved partly through whole exome and transcriptome sequencing of the tumor to detect various source mutations including single nucleotide variants, insertions, deletions, fusions, and others. It also requires accurate HLA typing from whole exome sequencing of the germline sample to derive the patient's HLA genotype for neoantigen predictions. This is followed by complex computational methods to characterize the neoantigens and synthesize the top candidates for vaccine development.
There are a number of parameters that can be considered in defining an immunogenic neoantigen and these are related to each step of neoantigen biology. Integrated bioinformatic pipelines and predictive models can provide information on peptide cleavage site, predicted binding affinity of the peptide to patient specific HLA alleles, binding stability, peptide presentation, expression abundance, clonality, alterations in the antigen presenting machinery, and others. There is also significant variability in these pipelines in terms of which parameters and neoantigen features scientists decide to include in the prioritized list of potential neoantigen candidates.
The Parker Institute brought together a team of researchers from academia, industry and non-profit centers working in the field of neoantigen characterization for therapy design. They formed the Tumor Neoantigen Selection Alliance (TESLA) with the objective of standardizing the process of accurately and reproducibly identifying neoantigens capable of eliciting an immune response, and creating a validated dataset that the community can use for effective benchmarking. TESLA designed a study for comparing key parameters to better predict neoepitope immunogenicity [Wells et al, 2020]. The study included six tumor and normal samples from melanoma and non-small cell lung cancer. All samples were sequenced and analyzed using a single source. The same genomic information which consisted of whole exome sequencing data from the tumor and normal, whole transcriptome sequencing data from the tumor, and HLA typing was provided to twenty-eight participants in this exercise. Participants were asked to generate a list of ranked neoantigens from this data using their own algorithms. TESLA then took the top neoantigens as defined by each participant, a list of ~600 neoantigens in total, synthesized these peptides, and performed in vitro testing to validate whether these selected neoantigens are able to bind to the patient's HLA alleles and form a multimer. They found that only 6% of the top ranked peptides were actually immunogenic and observed substantial diversity in the features used by each participant in defining what is an immunogenic neoantigen. They then looked at the overlapping features enriched in the immunogenic set of peptides and identified key essential metrics that defined immunogenicity.
From these elaborate studies, four key features were established in determining immunogenic potential. These features belong to two major categories:
Personalis® Neoantigen Prediction Model
Personalis® has been actively refining a machine-learning model to improve neoantigen identification and predictions. The model is very well aligned with the major findings from TESLA and has integrated the four key features into the design and pipeline output. These include all the presentation features as well as foreignness.
Our prediction algorithm, SHERPA™, is developed and trained on immunopeptidomics data. Briefly, to build our model, we created a training dataset starting with mono-allelic HLA transfections in empty cell lines to obtain the profile of peptides that actually bind to these alleles. Careful thought has been put into selecting these alleles to optimize both allelic diversity and population coverage and enable accurate and comprehensive modelling of peptide processing and presentation. We then purified the HLA/peptide complex, eluted the bound peptides, and performed mass spectrometry experiments to identify the peptide sequences. These sequences made up the Class I dataset used to train our algorithm. We created two models to recapitulate multiple parameters in the neoantigen biology. The SHERPA-Binding algorithm uses peptide sequence and HLA binding pocket information to predict a binding rank for affinity. The SHERPA-Presentation model takes this a step further, and expands by incorporating antigen processing machinery, gene expression information, and flanking regions to predict a more comprehensive presentation rank.
Personalis® has been supporting the personalized neoantigen-based cancer vaccine and cell therapy development for years and is at the fore-front of that process. The personalized workflow for customers begins by receiving a patient tumor biopsy and normal samples from a clinical site. The samples are sequenced and analyzed using the ImmunoID NeXT Platform® to produce data-rich analytics and putative neoantigen candidates from SHERPA. This information is then utilized by our partners developing personalized cancer vaccines or cell therapies to design a treatment tailored to the patient's immunogenomics profile.
In addition to SHERPA, ImmunoID NeXT™ provides deep sequencing, augmentation of all 20,000 genes, and high resolution to detect all sources of potential immunogenic neoantigens: SNVs, InDels, fusions, and others. The platform integrates HLA typing capabilities key for patient HLA allele-specific predictions. The advanced analytics related to alterations that impact HLA genes and the neoantigen processing and presentation machinery inform whether the predicted neoantigen on the tumor is visible to the immune system. The data from ImmunoID NeXT can also be applied to retrospective studies to better understand neoantigen biology, improve prioritization strategies, and explore biomarkers of response and mechanisms of resistance to the vaccine.