Skip to main content
Premium Trial:

Request an Annual Quote

MiRoncol Eyes Next Steps to Commercialize MicroRNA-Based Multi-Cancer Screening Test

Premium

NEW YORK – Molecular testing startup MiRoncol is seeking new investors to support continued validation and commercialization of a multi-cancer early detection (MCED) technology based on the analysis of circulating microRNA molecules.

The investigators behind the test published a study in late September describing their initial development of what turned out to be a four-miRNA assay, using machine learning to glean a cancer-specific signal from an initial case control cohort, and then validating it in three independent datasets that included 13 tumor types.

Based on those results, published online in Scientific Reports, the firm now hopes to build support for additional retrospective studies and eventual prospective validation, CEO Victoria Xu said this week.

Companies and research groups have taken a variety of approaches in recent years to build assays that can detect cancers in the blood of individuals who show no signs of disease. One of the most prominent, and the first to begin selling tests clinically, is Grail, whose Galleri assay is based on the detection of methylation patterns in circulating cell-free DNA (cfDNA) molecules.

Other examples, such as the Johns Hopkins' CancerSEEK assay — now being commercialized by Exact Sciences — have coupled multiple signals, including cfDNA methylation and proteins. Approaches have also been developed that analyze DNA fragmentation patterns.

Prior to Xu's investment and appointment as CEO, the scientific development of MiRoncol's miRNA-based take was spearheaded by Andrew Zhang, at the time a student at Del Norte High School in San Diego, and Hai Hu, CSO at the Chan Soon-Shiong Institute of Molecular Medicine.

Their strategy for gleaning a microRNA signature followed the model of a variety of others in the field, applying machine learning to an initial case-control cohort and pushing it to identify any discerning signals that differentiate cancer cases from matched healthy samples.

Zhang and Hu started with miRNA data from 1,408 cancer cases across seven tumor types obtained from the National Cancer Institute's Gene Expression Omnibus. These were paired with another 1,408 gender-matched non-cancer controls.

Bioinformatic analysis revealed four miRNAs — hsa-miR-5100, hsa-miR-1228-5p, hsa-miR-8073 and hsa-miR-663a — that together provided the best discrimination of cancer versus non-cancer. The researchers then tested this panel across three independent sample sets, also from the GEO, totaling 4,874 cancer cases and 13 cancer types.

The result was 90 percent sensitivity for nine tumor types (lung, biliary tract, bladder, colorectal, esophageal, gastric, glioma, pancreatic, and prostate cancers) and between 75 percent and 84 percent sensitivity for three other cancer types (sarcoma, liver, and ovarian cancers) — all at 99 percent specificity.

As other MCEDs have moved forward in their own development, a crucial question has been whether their performance is high across all tumor types, or only a subset, as well as whether the assays performed well for the earliest-stage cancers that clinicians hope to identify and not just late-stage tumors that offer little opportunity for a cure.

The MiRoncol dataset contained about 1,500 lung cancer cases and 1,205 gastric cancer cases. There were also 769 prostate cancers, but only about 300 or 400 examples from other tumor types like bladder, ovarian, liver, and esophageal cancers.

In terms of stage breakdown, 87 percent of the lung cancers were stage I or II, as well as all the gastric cancers. In contrast, only 70 percent of liver cancers, 66 percent of esophageal cancers, and 35 percent of ovarian cancers were early stage.

Zhang and Hu reported that sensitivity for early-stage lung cancers was a striking 99 percent, but case-level stage information wasn't available for each of the other tumor types. That said, they were able to calculate some stage-specific numbers for these other cancers.

For example, all the gastric cancer patients were stage I or II, thus, the 100 percent sensitivity seen in the validation implies 100 percent sensitivity for early-stage tumors. Similarly, 88 percent of bladder cancer cases and 93 percent of the prostate cancer cohort had node-negative disease.

"With 99 percent and 98 percent sensitivity for these two cancers, the sensitivity for stage I or II bladder and prostate cancers should be very high, as well," the authors wrote.

And finally, based on the prevalence of early-stage cases in the esophageal and liver cancer groups, it would be "reasonable to speculate that the sensitivity for stage I or II of these two cancers should not be far off from the 92 percent and 84 percent sensitivity reported for all stages," they added.

Zhang and Hu noted that the original studies that generated the miRNA data in the GEO database proposed their own cancer-predictive miRNA panels, with only one miRNA overlapping more than a single cancer type.

Even though some of those studies showed higher performance than the MiRoncol panel, they calculated that the cumulative incidence of false positives would be as much as 33 percent if the various tumor-specific panels were to be applied as a group.

The question for MiRoncol now is whether this performance will hold up in future validation studies in new independent cohorts. As the data stands, the miRNA panel's sensitivity for early-stage tumors rests leaps and bounds above what other assays, like Grail and CancerSEEK, have been able to demonstrate, but this will have to be proven out beyond an initial computational analysis.

Zhang and Hu also raised the question of cost in their study conclusion, as did Xu in an interview this week. Methylation and fragmentation-based approaches require comparatively expensive next-generation sequencing technologies, while a four-miRNA panel could be run on ubiquitous and relatively cheap quantitative PCR systems.

This also suggests an advantage in testing scope and accessibility, with the potential for disseminated IVD tests rather than a centralized laboratory developed test model. According to Xu, this "democratization" potential was a particular driver for her interest in the technology and forming the company.

Zhang and Hu wrote that with the current study limited to computational analysis using public data, the results shouldn't be viewed as definitive. The path toward a commercial test is a long one, Xu added.

Nevertheless, the next steps are well defined, she said, with other companies like Grail having laid a trail that MiRoncol could mirror if it can secure the necessary funding.

Being behind these other firms isn't something that she is worried about, Xu said. The company doesn't see them as competitors as much as peers, with shared goals to reduce the impact of cancer on human lives and livelihoods.

"We are all in this together creating this new era of early detection," she said.

Grail and others have faced significant epidemiological skepticism regarding the likelihood that their tests, even if they do produce a shift in cancer diagnoses to earlier stages, will meaningfully affect patients' overall outcomes. The answer to those questions won't come until true clinical utility data can be borne out from prospective studies.

Xu said that in the meantime, health economic modeling and positive clinical impact data from existing cancer screening modalities may also be applicable to MCED tests, especially low-cost ones with high performance in early-stage cases.

Xu, herself an investor in MiRoncol, didn't detail how much the Philadelphia-based company has raised thus far, but described the firm as still very early in its commercial evolution and looking for backing to support its next steps, which include independent validation, ideally across a few thousand samples. The company filed a patent application regarding its technology, with Hu and Zhang listed as inventors in December 2023.