Together, we can transform the future of cancer care.
Tracking early stage breast cancer with ultrasensitive ctDNA
Information for patients with breast cancer
Monitoring therapy response with ultrasensitive ctDNA across cancer types
Ultrasensitive ctDNA detection in early-stage lung cancer
Personalis is transforming the development of next-generation therapies.
Neoantigen discovery and assessment are critical for the development of personalized cancer therapies and neoantigen-based biomarkers, requiring comprehensive detection of tumor-specific genomic variants and accurate prediction of MHC presentation of neoepitopes originating from such variants. Personalis delivers a comprehensive survey of putative neoantigens by combining highly sensitive exome-scale DNA and RNA sequencing with our NeoantigenID analytics.
The NeoantigenID workflow begins with your tumor and normal specimens and proceeds as shown in Figure 1:
Accurate neoantigen prediction with SHERPA enables you to determine candidate neoantigens for rapid development of personalized cancer therapies, as well as facilitating the generation of neoantigen burden-based composite biomarkers such as NEOantigen Presentation Score (NEOPS) that can potentially better predict response to immunotherapies compared to other biomarkers, such as TMB. Accurate neoantigen prediction with SHERPA enables you to identify, assess and prioritize candidate neoantigens for rapid development of personalized cancer therapies, as well as facilitate the generation of neoantigen burden-based composite biomarkers such as NEOantigen Presentation Score (NEOPS). NEOPS integrates neoantigen assessment data provided by SHERPA with identified somatic variants in antigen presenting machinery and HLA LOH, as determined via our proprietary DASH algorithm, in order to provide the potential to better predict response to immunotherapies than using TMB alone.
Figure 1: NeoantigenID
HLA binding is currently the most well-established criteria for ranking neoantigen candidates. We’ve leveraged recent advances in training data generated from mass spectrometry to provide you with a larger dataset of peptide binders and non-binders for individual HLA alleles. This new binding data takes two important additional components into consideration: cleavage and transportation, both critically important for presentation assessment. These advancements power our Systematic HLA Epitope Ranking Pan Algorithm (SHERPA), our pan-predictive machine learning model for predicting MHC class I presentation.
SHERPA utilizes proprietary, high quality immunopeptidomics data, publicly available & curated mono- and multi-allelic data, as well as binding affinity data as a training set (Figure 5). Publicly available multi-allelic data from several tissue types were systematically reprocessed and deconvoluted to capture the diverse facets of antigen processing and presentation. The integration of different training data types results in decreased bias, increased generalizability, and improved performance of SHERPA.
Figure 2: SHERPA: Neoantigen Machine Learning Algorithm using Proprietary Engineered Cell lines & Mass Spec Data
Multiple modeling strategies were combined to accurately predict neoantigens for all known alleles. The SHERPA-Binding algorithm uses both the peptide and binding pocket information to predict a binding rank. The SHERPA-Presentation algorithm incorporates additional, critical features such as expression level of the source protein, proteasomal cleavage, and gene propensity to predict a more comprehensive presentation rank (Figure 6).
Figure 3: SHERPA-Binding and SHERPA-Presentation Prediction Models
We evaluated SHERPA performance on ~10% held-out mono-allelic data set, mixed with negative examples in a 1:999 ratio (commonly assumed prevalence). The precision-recall curves demonstrate that SHERPA models have consistently higher precision at all recall values compared to other publicly available prediction algorithms (Figure 7A). Both SHERPA models also have better positive predictive value (PPV) compared to publicly available prediction tools (Figure 7B). SHERPA-Presentation has a better PPV compared to SHERPA-Binding model, attesting to the utility of presentation-specific features. Boxplots in Figure 7B denote the distributions of PPVs (top 0.1%) across alleles within the mono-allelic immunopeptidomics held-out test data. Distributions are shown to compare SHERPA with other publicly available models.
In addition to its status as an emerging biomarker of interest in the era of cancer immunotherapy, HLA genotyping is also an essential component of the neoantigen prediction process. Personalis’ HLA typing tool, HLA-Map, has been integrated into NeoantigenID; enabling the highly-accurate in silico typing of all HLA Class I and Class II loci, which is critical for ensuring the precision of downstream peptide-MHC-binding predictions.
Table 1: HLA-Map’s HLA genotyping performance for both HLA Class I and Class II loci.
To confirm the accuracy of HLA-Map, we performed a comprehensive analytical validation study. This validation study was performed on a total of 15 proficiency testing samples with known, but blinded HLA genotype profiles. Ten of these samples were sourced from the American Society of Histocompatibility and Immunogenetics (ASHI) and five additional samples were obtained from the College of American Pathologists (CAP). Each of these samples had previously been independently genotyped via various orthogonal clinical tests, and these results against which our own results were compared. As is demonstrated in the table below, HLA-Map performed exceptionally well in accurately genotyping not only the HLA Class I loci, but also the more challenging HLA Class II loci.
The success of immune checkpoint inhibitors have revolutionized cancer treatment. However, the fact that the majority of cancer patients do not respond favorably to such immunotherapies has resulted in an explosion in the breadth of research efforts to identify new biomarkers of response and/or resistance to these new classes of cancer therapeutics.
Given that the mechanism of action of these therapies is contingent on the dynamic interplay between the tumor and the host’s immune system, the role of the antigen processing machinery (APM) in ensuring that tumor-specific neoantigens are successfully presented to the adaptive immune cells has garnered increasing attention in the search for more effective biomarkers. More specifically, loss of heterozygosity (LOH) impacting the HLA Class I genes has emerged as a means by which solid tumors can evade immunosurveillance by reducing the repertoire of neoantigens that can be presented to the immune system, and this phenomenon is now recognized as a key resistance mechanism to immune checkpoint blockade or, ICB (McGranahan et al., 2017; Tran et al., 2016).
To provide our partners with the most comprehensive cancer immunogenomics platform, Personalis developed DASH (Deletion of Allele-Specific HLAs): a machine-learning-based tool to capture the unique features associated with each individual HLA Class I region. When combined with our ACE-augmented sequencing data, we are able to accurately assess HLA LOH.
In order to validate our performance, we assessed the limit of detection (LOD) of DASH using three tumor-normal cell line pairs with HLA LOH in at least one locus. We sub-sampled the tumor sequencing data and mixed it with complementary normal sequencing data to achieve simulated purity levels. Next, we mixed the HLA-mapping reads across a range of ratios to simulate the potential spectrum of tumor purities and sub-clonalities. Both LOH-HLA and DASH have nearly perfect specificity (>99%, data not shown) across tumor purities and sub-clonalities.
For fully clonal HLA LOH events, consistent sensitivity is achieved with >25% tumor purity for both algorithms. However, DASH has significantly higher sensitivity to detect sub-clonal events than LOHHLA (Figure 8).
Additional validation studies utilizing several novel, orthogonal methods have been completed and the results of these studies can be found here.
Figure 7: DASH has significantly better sensitivity than LOHHLA for sub-clonal HLA LOH events.
