NEW YORK – With promising new data published this week, proteomic cancer testing firm Novelna believes it can develop a blood-based multi-cancer screening test at a fraction of the cost of DNA-based competitors with dramatically improved sensitivity to boot.
With this combination of pricing and performance, the firm hopes to more easily sell regulators and payors on a new vision of population-wide screening in asymptomatic individuals — a challenge that has become widely debated in regard to the future prospects of emerging DNA-based tests like Grail's Galleri.
"There is a high cost that make these tests only affordable for a specific population and in specific countries," Novelna founder and CEO Ashkan Afshin said in an interview. "With the goal of increasing the sensitivity and also reducing the cost, we saw a great potential in proteomics."
The genesis of the company's approach was formed in conversation with Bogdan Budnik, a principal scientist at Harvard's Wyss Institute and now head of proteoformics at Novelna. Afshin said that Budnik has extensive experience in the field of mass spectrometry, but while recent improvements have begun to allow higher-throughput mass spec analyses of thousands of proteins, there remain challenges, especially in measuring lower-abundance molecules. With that in mind, the Novelna team turned to a proximity-based methodology marketed by Olink, a Swedish firm expected to soon be acquired by Thermo Fisher Scientific.
The company's new study — published online in the journal BMJ Oncology this week — appears to support that strategy, which yielded two 10-protein sex-specific cancer detection panels with sensitivity for early-stage cancers more than double what has been reported for DNA-based MCEDs at a maximized 99 percent specificity.
Led by Budnik, the study analyzed plasma samples from a cohort of 440 individuals, including those diagnosed with 18 types of early-stage solid tumors, plus healthy controls.
Using Olink's Explore 3072 library, investigators measured more than 3,000 high-abundance and low-abundance proteins in each sample, using a multistep statistical approach to identify two sex-specific panels of proteins that could discriminate the cancers from controls. An additional set of proteins was also able to determine the location of cancer in the body.
The two sex-specific cancer detection panels were notably accurate. For stage I tumors, with a specificity set at 99 percent, the tests could identify 92 percent of cancers in males and 84 percent in females. The 150-protein location panel correctly identified a tumor location in more than 80 percent of the cohort's cancer cases.
"One challenge [with DNA-based testing] is that many tumors may not release DNA, or by the time that DNA is released into circulation, it will be too late," said Afshin. Circulating proteins, hypothetically, might capture signaling associated with the first steps of tumor formation, long before a cancer reaches a state where it sheds any DNA.
Interestingly, a majority of the proteins Novelna identified as cancer-detectors were present at relatively low levels, which he said is further evidence that the signal the company is picking up reflects early-stage cancer development. Likely, some of this signal is from the tumor itself and some from the immune response to tumor development, "so that's another reason that we believe in this hypothesis from the biology side, as well," Afshin said.
In a commentary appearing alongside the Novelna study, Ohio State University professor Holli Loomans-Kropp wrote that the firm's sex-specific approach could be another factor influencing the observed boost in sensitivity. For one, she wrote, women have higher rates of invasive cancers at earlier ages than men, and the cancers affecting women and men differ in both their type and frequency.
There are also genetic differences that affect sex-specific aspects of cancer incidence. For example, Loomans-Kropp wrote, "Germline TP53 mutations are associated with increased cancer incidence and earlier onset age in women compared with men … [and] other sex-specific differences have been observed for overall mutational burden, copy number alterations, and methylation patterns."
"Although several problems need to be addressed before MCEDs can be deployed at a population scale, a method to improve on current issues of sensitivity and specificity may be use of sex-specific detection panels," she added.
This appeared to bear out in the Novelna team's Olink results. When the team tried to do a sex-agnostic comparison between cancer and normal samples, the correlation was poor: 80 percent of proteins with a p value below 0.05 in the male cohort had no significant difference in females, with the same true vice versa.
The authors said that in males, the 10 proteins with the lowest p values differentiating cancer from normal were IGFBP2, CKB, CEACAM16, MMP1, ENPP5, ELN, ITIH3, WNT9A, KRT19, and SSC4D. In females, the 10 proteins with the lowest p values were C3, SEMA4C, NAPRT, GBP2, MAEA, SEPTIN9, PSMD5, IL1A, CLTA, and CNST.
Although the performance of the Novelna panels varied across cancers, average sensitivity didn't drop nearly as low as has been seen in data from DNA-based MCED trials, even for the most challenging tumor types such as bladder cancer in females and thyroid cancer in males.
Other findings included the fact noted by Afshin that low-abundance proteins appear to be preferentially represented in cancer-associated signaling. According to the authors, nearly all the proteins included in their predictive panels were in the low-concentration part of the human plasma proteome. The data also suggested that down-regulated proteins are crucial for discriminating different cancer types with only a few tumor types able to be uniquely identified by upregulated proteins.
The authors wrote that this may be due to the fact that "most of our knowledge about the role of proteins in cancer pathways comes from studies of transcriptome at the tissue level in advanced stages of cancer, and the expression of proteins at the mRNA and protein levels do not always correspond."
Further validation will be required of the improved sensitivity that the Novelna team saw in their current study, Afshin said.
The retrospective cohort used for this proof of concept was relatively small and had limited representation in terms of different stages across all the 18 tumor types, as well as limited information on patient comorbidities and other clinical complexities. That said, Afshin argued that even a 10 percent dip in performance in future validation work would still leave the Novelna tests' sensitivity above that of DNA-based MCED tests.
With only 10 proteins each, the company's two sex-specific detection panels would be amenable to translation to enzyme-linked immunosorbent assay, which offers an affordable broadly deployable platform that reflects the company's goal to enable true worldwide population-scale MCED screening.
With 150 proteins, the tumor localization assay would need to use a different technology, but if it were used as a reflex test — only performed in patients who have a cancer detected by the initial assays — it would be in a tiny fraction of that global population, dramatically lessening the impact of a potentially higher cost.
The company, based in Palo Alto, California, is currently working on raising money to support an independent validation in a larger, more comprehensive cohort, including representation of different geographies and ethnicities, and enough cases to be able to assess any effect on the tests from patient comorbidities.
Afshin said the firm is targeting governmental and nongovernmental funding, as well as private investors. "Basically, we need to have two other validation studies, and then the test will be ready for commercialization … [and] our goal is to ensure that we complete those two studies over the next two to three years," he said.