Q&A with John West, Founding CEO of Personalis
Personalis has been building and innovating for more than a decade, but we believe we’ve only just begun. In this multipart series, we have been talking to Personalis’ founder and CEO, John West, about this journey and where we’re headed, both as a company and industry.
In a previous post, we discussed how Personalis has pioneered in the field of molecular residual disease (MRD), developing a whole genome tumor-informed approach to achieve tumor detection limits down to one part per million. Here we discuss how such sensitivity changes the way we think about cancer management and what’s next.
Q: We’ve said that the extraordinary MRD sensitivity of NeXT Personal™ changes the way we think about cancer management. Why is that?
A: Early MRD tests were an advance in that they could detect cancer recurrence before traditional standard of care imaging. On the other hand, the limited sensitivity of these assays often meant that patients would be tested again and again, until finally the cancer progressed up to their limits of detection. Getting to that point of detection became the goal of many studies.
At Personalis, our goal has been to detect every recurring cancer at the first time point after surgical resection. While we have not yet reached 100%, that goal does change how we think about the tests we develop. If residual disease is detected right after surgery, our job is not done. It shifts our focus to the clinical choices that an oncologist will face and how we can help.
Let me describe three of these choices:
First, most tumor-informed MRD assays don’t provide DNA sequencing coverage for drug-targetable variants. Oncologists need this tumor characterization information, in addition to the ongoing quantification provided by an MRD assay, to treat their patients. For instance, a growing list of therapies have been FDA approved for adjuvant use, but many are designed to target specific mutations. The efficacy of osimertinib (Tagrisso), for example, depends on the presence of a somatic EGFR mutation. The efficacy of olaparib (Lynparza) depends on the presence of a germline BRCA mutation. So, the question is whether the tumor has a mutation targetable by an approved therapy. And, if so, whether a therapy is available to target that particular somatic mutation in a patient’s tumor. Therefore, it will be important to monitor the allele frequency of that variant in parallel with tumor quantification. If the variant is just in a subclone of the tumor, its allele frequency may decline more than the overall quantification of the tumor, showing the effect of the targeted therapy just on the subclone.
Second, most targeted therapies eventually fail, as resistance mutations emerge. The recent PADA-1 trial in breast cancer showed that progression-free survival can more than double if resistance mutations are detected in the ESR1 gene in cell free DNA from the blood plasma and used as the basis for changing therapy. Most tumor-informed MRD assays don’t provide DNA sequencing coverage for these well-known resistance variants. But oncologists need this tumor characterization information, in addition to the ongoing quantification provided by an MRD assay, to treat their patients.
Figure 1: Resistance mutations that emerge in the ESR1 gene in breast cancer patients receiving an aromatase inhibitor as adjuvant therapy (J.T.Lei, Baylor College of Medicine, J. Cancer Metastasis Treatment, May 15, 2019). The Pfizer PADA-1 trial showed that if, on detection of these, therapy is switched to fulvestrant, progression-free survival more than doubled.
Third, we know that not all cancers that appear after surgical resection are recurrences. A major NCI study found that 19% of cancers diagnosed in the United States are the patient’s second or subsequent primary. The longer a patient survives an initial cancer, the more time there is for a second cancer to arise. Because these subsequent tumors are not genetically related to the first, most tumor-informed MRD assays are not designed to detect them. In addition, a second primary will generally not harbor the same targetable mutations of the first one, but it may have its own mutational signature. Because some driver mutations (e.g., KRAS G12C) occur much more frequently in some cancer types than others, they can also help narrow the search for where a second primary may be located. Again, oncologists need this tumor characterization information, in addition to the ongoing quantification provided by an MRD assay, to treat their patients.
Figure 2: As more cancer patients are surviving, and for longer, there are more and more cases of subsequent cancers which are not recurrences. Tests to detect cancer after surgical resection, which do not detect these second cancers, should be considered false negatives. Source: Personalis.
As MRD tests become more sensitive, they will be used at fewer time points to detect recurrence, and at more time points to manage the recurrences. High MRD sensitivity is required to quantify the growth or shrinkage of a tumor as it is being managed, but the examples above show that is not enough. Oncologists also need information characterizing each tumor over time – to select an initial therapy, to guide changes in therapy if resistance develops, and to detect and manage second primaries.
Q: How does this change the way that tumor monitoring assays need to be designed?
A: Tumor monitoring assays of the future will need to combine tumor-informed content and content from databases. NeXT Personal takes this approach. Tumor-informed content (polymorphisms identified by sequencing a sample from the patient) is important to achieve high MRD sensitivity. And, the databases we use, in addition to tumor-informed content, include:
- Cancer-related germline variants (like the BRCA variants mentioned above)
- Somatic variants, which can be targeted by FDA-approved cancer drugs (like the EGFR mutations mentioned above)
- Variants known to emerge in resistance to specific drugs (like the ESR1 mutations mentioned above)
- Common driver mutations, which may emerge indicating the presence of a second primary (like the KRAS mutation mentioned above)
By combining tumor-informed and database-derived content in a single assay, we make the most of very limited sample amounts. We plan to validate our NeXT Personal MRD test down to 5 nanograms of input. By leveraging almost two thousand positions across the genome, the MRD part of NeXT Personal is designed to stay sensitive, even at these low levels. By supplementing this tumor-informed content with thousands of loci from a broad set of cancer-related databases, we can also provide critical characterization of a tumor, all from the same 5 ng input.
Figure 3: Many driver mutations come from one or a small number of source organs. Example here: Sanger Centre COSMIC database for EGFR L858R. In addition to its high MRD sensitivity, NeXT Personal targets thousands of loci from a database of common cancer mutations, enabling it to potentially detect an otherwise unseen second primary and provide a good idea where it is.
Q: You have outlined this evolution in thinking for early-stage disease, following surgical resection. Does the same apply to patients diagnosed with metastatic cancer?
A: Absolutely. As mentioned above, we expect increasingly sensitive MRD tests to be used at fewer time points to detect recurrence, and at more time points to manage the recurrences. Patients diagnosed with unresectable metastatic cancer need both tumor quantification and tumor characterization, right from the first time point. The more advanced a cancer case is, the more important it is to provide both.
As a metastatic patient progresses, there may be resistance to providing multiple tubes of blood for testing. If 5 ng of cfDNA is all that is available at a late-stage time point, we need to be able to provide all of the data needed, from a single integrated assay.
Personalis has been looking to the future, beyond “just MRD,” with a focus on building tomorrow’s gold standard. Ultimately, the company aims to improve patient outcomes with more insightful diagnostics and faster pharmaceutical development.