NEW YORK – Presentations at last week's annual meeting of the American Association for Cancer Research suggest that there are still numerous stones unturned in terms of methods and biomarker targets for multi-cancer early detection (MCED).
"The more we … identify different alterations that occur in cancer, what we've seen is that there are various alterations [including] genomic changes, chromatin changes, nucleosome occupancy, repeat landscapes, epigenetics, gene expression, and all of these [that] can ultimately be detected in the blood," Johns Hopkins University professor Victor Velculescu said during a session discussing some of the latest discoveries.
"It's an exciting time to think about how we incorporate these in methodologies as we continue characterizing or developing new technologies," he added.
Presentations at the conference included studies of DNA fragmentation patterns, the genomic repeat landscape, and novel proteomic targets, among others.
A team from China shared data on a test it has developed called MCTarg, which harnesses blood-based metabolomic signatures, honed by machine learning, to screen for lung, gastric, and colorectal cancers.
These investigators, working with a company called Metanotitia, enrolled 951 cancer patients and 889 individuals representing both healthy controls and those with potentially confounding benign diseases across three medical centers. They tested plasma samples using mass spectrometry, dividing the overall cohort into a discovery set and an internal validation set.
The team also performed an external validation in an independent cohort of 108 cancer patients and 125 healthy individuals from two other centers.
Using data from the training set, the researchers were able to create two screening models, one for a low-risk population (MCTarg-1) and another for higher-risk individuals (MCTarg-2). In a poster presentation, scientists from the firm reported nearly 99 percent sensitivity at 98 percent specificity for MCTarg-1 in an internal validation cohort, and 94 percent sensitivity at 95 percent specificity in the external validation cohort.
MCTarg-2 for high-risk populations demonstrated 65 percent sensitivity at 86 percent specificity externally. Notably, sensitivity for early-stage cancers — stage I and II tumors that have the best chance of a cure — was higher than what has been seen for some other multi-cancer detection assays at 79 percent for MCTarg-1 and 69 percent for MCTarg-2.
After locking in a further refined biomarker panel including 66 metabolites, investigators reported that MCTarg-1 achieved 73 percent sensitivity at 87 percent specificity in the external validation cohort, while MCTarg-2 had 57 percent sensitivity at 84 percent specificity.
MCTarg-2 also showed 69.4 percent sensitivity at 91.7 percent specificity and 57.4 percent sensitivity at 84.0 percent specificity, respectively.
With further large-scale validation and the inclusion of additional cancer types, MCTarg has the potential to become a universally applicable, simple, and cost-effective method, enabling early detection and localization of common cancers in large populations, the researchers noted.
Indian diagnostics firm Strand Life Sciences also presented on a multi-cancer detection approach using targeted methylation, similar to platforms being advanced by US companies like Grail, Guardant Health, and Exact Sciences.
Strand developed its test, called CancerSpot, in a cohort of 688 cancer patients across 10 tumor types and 326 healthy controls representing 10 cancer types, using machine-learning models to assess potentially cancer-linked methylation signatures from across approximately 7,000 genomic targets.
According to company researchers, the resulting discriminator showed 79 percent sensitivity and 97 percent specificity for stage I-III tumors. It was also able to identify the tumor tissue of origin within its top two predictions about 77 percent of the time.
Similarly to Metanotitia's assay, CancerSpot maintained sensitivity in the high 70 percent range even for early-stage tumors.
During a session on advances in cancer interception, Kathleen Burns, chair of the department of pathology at the Dana-Farber Cancer Institute, also discussed the promise of a specific protein biomarker to aid early detection.
Protein biomarkers have previously shown limited sensitivity, specificity, or both for cancer screening purposes. But Burns and her colleagues have been studying a specific molecule, long interspersed element-1 (LINE-1, L1) open reading frame 1 protein (ORF1p), which they have shown to be overexpressed in carcinomas and high-risk precursors during carcinogenesis with negligible detectable expression in corresponding normal tissues.
Burns and her colleagues have engineered ultrasensitive digital immunoassays that detect ORF1p concentrations in patient plasma samples across multiple cancers with high specificity and are now working to optimize their technology for further clinical research.
Velculscu also highlighted recent developments in the genomic space, including a method developed by Johns Hopkins researcher Daniel Bruhm called GEMINI that allows for the detection of cancer-associated mutations using low-coverage single-molecule sequencing. The approach takes advantage of the fact that mutations occur more frequently in certain genomic regions, and these patterns are different in cancer versus normal cells.
Velculescu is also a founder of cancer early detection firm Delfi, and he said that he and his team at the company are "looking forward to potentially incorporating this approach."
He also cited another new methodology that focuses on the repeat landscape, a part of the "dark matter of the genome" that Velculescu said has only recently been able to be interrogated.
New efforts have been able to quantify specific repeat families as they are altered in cancer and detect these patterns in cell-free DNA.
In the end, Velculescu said, he and his colleagues believe that most of these various types of changes ultimately reflect in DNA fragmentation patterns, but it is becoming clear, he said, that there is orthogonal, or added value, in different approaches.