NEW YORK – Precision medicine and population genomics firm Personalis has added a new capability to its sequencing and informatics service that it hopes will support the development and implementation of a new class of composite biomarkers to predict patients' responses to cancer immunotherapy.
Announced last week, the firm's new SHERPA (Systematic HLA Epitope Ranking Pan Algorithm) technology is available to customers of Personalis' NeXT platform, which combines exome-wide analysis of DNA and RNA with interrogation of multiple tumor microenvironment and immune cell components from a single preserved tumor tissue sample.
Alongside SHERPA, Personalis also debuted an immunotherapy-predictive assay it developed using the epitope-ranking algorithm, which it calls NEOPS (Neoantigen Presentation Score).
The predictor combines SHERPA-derived neoantigen predictions with molecular signals of tumor immune evasion to create a composite biomarker that the company believes could be more effective in predicting immunotherapy response than established markers like PD-L1 status and tumor mutational burden (TMB).
"Identifying effective biomarkers for cancer immunotherapy has proven challenging due to the complex biological mechanisms underlying patient response. We believe that more sophisticated, multimodal biomarker models like NEOPS represent the next step in this pursuit. By combining accurate tumor neoantigen prediction with mechanisms of immune evasion, we can create a more predictive model of therapy response," Richard Chen, Personalis' chief scientific officer, said in a statement.
Personalis CEO John West added that the company sees direct potential for NEOPS as a clinical companion diagnostic, with pharmaceutical customers already investigating the score in ongoing trials. The firm also envisions broader potential for the core NEOPS signature (and the larger SHERPA platform) to serve as a jumping-off point for even more informative drug- or indication-specific tools.
"If you're a pharmaceutical company and you have a particular new type of immunotherapy, you may want the information [from NEOPS]. But … we also see a number of groups, including some of our customers, who use RNA data — because we're looking at all 20,000 genes of RNA as well — to develop gene expression signatures that that reflect the activity that may be relevant to the drug, and they may want to combine that with the new antigen results out of SHERPA," West said.
He further noted that NEOPS is an example of a composite biomarker with clinical potential.
"As we think about cancer drugs of the future and the companion diagnostics that they require … I think this is where biomarkers are headed," West said.
The company is not alone in this view, with researchers and test developers increasingly turning to composite, or multi-modal approaches to overcome the limitations of current tools like TMB, microsatellite instability, and PD-L1.
Personalis has already begun to share early data on NEOPS that the firm believes bodes well for its potential clinical translation.
In a presentation at the AACR virtual meeting in June, investigators reported on a study of samples from 55 late-stage, unresectable melanoma patients, in which a SHERPA-derived score significantly outperformed tumor mutational burden in predicting response to anti-PD-L1 therapy. It also better predicted progression-free survival.
The researchers also compared their neoantigen and resistance mechanism score with a straight read of neoantigen burden, demonstrating that the SHERPA-based approach was superior to what might be achievable using other neoantigen prediction platforms.
Neoepitope prediction is something Personalis' customers and collaborators had already been working on prior to the launch of SHERPA, and something that can be done with other comprehensive genomic platforms. But West said that the company's new tool is unique in that it enhances the identification of likely neoantigens based on incorporation of information about immune evasion mechanisms that the company built from the ground up through a comprehensive analysis of cell lines.
"The overall goal here is to be able to understand which mutations in a person's tumor might actually spur an immune response," West said. "You can have mutations in any gene … and the mutated DNA can lead to a mutated protein … but it isn't just a question of which mutations you have, but are they expressed or not? … and then will those peptides be displayed on the outside of the cell?"
According to West, the answer to these questions rests not only on factors related to HLA and MHC-complexes, which other neoantigen prediction tools might incorporate, but also a complex system of processing machinery inside of cells that also contributes to which peptides make it through the digestion process in such a way that they can be bound and presented.
"If they don't make it through that process, it doesn't matter if they would bind to HLA because they don't get to the point where they can even do that," he argued.
To address this, Personalis set out to comprehensively engineer and analyze human cell lines that express a single HLA allele of interest, analyzing patterns of resulting peptide binding using mass spectrometry and finding a way to incorporate the resulting knowledge into the prediction of which gene mutations are likely to result in proteins that actually become visible to the immune system.
"We started on this over four years ago, genetically engineering the cell lines, getting the mass spectrometry to work and then developing the machine learning algorithms to take advantage of that," West said.
With the NEOPS predictor that the company used SHERPA to develop, yet another layer is added: a readout of molecular signals associated with tumor escape mechanisms that West said can essentially "cripple the ability of a cell to be able to display these new antigens," regardless of the intrinsic transcriptional and cell-processing mechanisms in play.
Moving forward, West said that Personalis has begun to have more conversations with pharmaceutical companies about NEOPS. But he said one of the key elements to potentially advancing the composite method in dedicated companion diagnostic partnerships is FDA approval.
"We've been discussing, I think for the last six months or so, that we've begun the process to take our NeXT platform through the FDA for a single-site PMA. So that process is continuing. And that certainly has increased interest [from] pharma," he said.
According to West, immunotherapy prediction and personalization is where Personalis sees the greatest immediate benefit of its platform, so that's where it is focusing its attention right now. Beyond diagnostics, the company also sees potential for SHERPA to inform new precision drug development via identification of targetable mechanisms of immune evasion.
For example, patients in the AACR melanoma cohort who had a high composite neoantigen score and TCR clonality showed a variety of potential resistance mechanisms, including high expression of IDO1 or CTLA4, which may facilitate PD-1-independent immune escape, and mutations in the antigen presentation machinery that could lead to loss of cell-surface protein expression.
The more accurate neoantigen prediction that the system offers could also aid long-term efforts in personalized cancer vaccine development, West added, if other technological hurdles can be overcome in that realm.