Skip to main content

AACR Project GENIE Enables Analysis of Growing Actionability, Disparities in Precision Oncology

Premium

NEW YORK – At the American Association for Cancer Research's annual meeting, investigators demonstrated the ability of real-world data to track the increasing "actionability" of tumor biomarkers, explore treatment patterns and outcomes among patients without targetable alterations, and hunt for molecular features that may contribute to worse outcomes among non-white cancer patients.

The investigators conducting these studies drew on real-world data collected over five years within the AACR's Project GENIE, which stands for Genomics Evidence Neoplasia Information Exchange. The international data-sharing effort, launched in 2015, sought to amass the clinical and genomic data of cancer patients across institutions and use it to improve precision oncology. The project started with data submissions from eight academic institutions, but now contains patients' data from 18 institutions in North America and Europe.

GENIE publicly releases data every six months and is on its ninth release, which contains data from around 104,000 patients with more than 100 major types of cancer, and information on nearly 113,000 sequenced tumors.

With each data release, the genomic and clinical content with GENIE has grown, reflecting the evolution of precision oncology as a discipline and its increasing impact on cancer patients. The amount of "actionable" somatic mutations — according to the criteria Memorial Sloan Kettering has devised within its OncoKB database — has more than doubled within GENIE between its first and ninth iterations, according to Trevor Pugh, a senior scientist at the Princess Margaret Cancer Center in Toronto. 

Pugh attributed this growth in actionability to the increasing number of US Food and Drug Administration-approved precision oncology drugs in recent years. However, GENIE's ability to capture this, he said, shows its ability to "take the pulse of clinical genomics at a given time."

Within expanding datasets, researchers can also identify and learn about exceedingly rare tumor markers, which in turn, can provide investigators better insight into whether they should open new biomarker-defined treatment arms within precision oncology basket and umbrella trials or retire arms unlikely to enroll. "The strength of GENIE is in more cases," Pugh said. "There's a lot of value in continuing to grow the dataset and looking at that long tail of less frequent alterations."

Among GENIE participants, patients are being tested on larger NGS panels, he said, and investigators are also thinking about how to bring in additional data types, such as WGS, transcriptomics, and ctDNA. "RNA sequencing is starting to be deployed more as a clinical test," said Pugh. "From the analysis working group side, we would love to have this type of data. It's going to be enormously powerful [in answering] all sorts of questions, especially in and around immunotherapy." Investigators are exploring how to provide access to raw testing data from GENIE contributors, which would facilitate analysis currently not possible with summary-level data.

Project GENIE Chair Philippe Bedard, an oncologist at the Princess Margaret Cancer Center, noted that many other institutions have said they want to join GENIE. "It is a question of resources in terms of how many sites we can realistically take on … and continue to generate high-level data," he said.

Measuring actionability

Although the FDA is approving more and more biomarker-informed cancer drugs each year, testing often identifies tumor alterations for which there are no approved treatments. In such cases oncologists have to scan the literature, databases, and other resources to figure out whether there is sufficient evidence supporting the use of the biomarker to prescribe off-label or investigational therapies. 

According to Sarah Suehnholz, senior scientist for Memorial Sloan Kettering's OncoKB, this resource is particularly useful in this regard. OncoKB is a catalog, updated monthly, of the therapeutic and prognostic relevance of genomic alterations MSK has detected in patients' tumors. The knowledgebase currently contains information on 682 genes, more than 5,600 alterations, nearly 120 cancer types, and 95 drugs.

Oncologists can look up mutations found in their patients' tumors in OncoKB and explore the evidence supporting their use in treatment decisions. Biomarkers recognized by the FDA alongside an approved drug or recommended by professional guidelines bodies have level 1 or 2 evidence and are considered standard of care. The biomarkers in 3A and subsequent levels are considered investigational and have increasingly sparse evidence.

At AACR, Suehnholz shared analysis showing how the actionability of biomarkers in OncoKB has changed over time and how that advancing knowledge then proliferates and impacts the broader cancer community represented in GENIE. In her analysis, Suehnholz used the February 2018 and March 2021 versions of OncoKB to track how the "actionability landscape" has changed for samples currently in GENIE.

Looking at level 1 and 2 OncoKB evidence categories, 56.7 percent of samples in GENIE would have had an "actionable" cancer biomarker in 2018, versus 61.1 percent in 2021. During this time, there was a 7 percent increase in samples with a level 1, FDA-approved biomarker, and a 3 percent decrease in samples with "investigational" level 3A biomarkers. This "demonstrates that the field of precision oncology is dynamic and its utility continues to grow," said Suehnholz.

During this time there was also a "dramatic shift" in the actionability of biomarkers for certain tumor types due to FDA approval of several biomarker-informed drugs. Additionally, three tissue-agnostic approvals — for pembrolizumab (Merck's Keytruda) using microsatellite instability status and tumor mutational burden, and for larotrectinib (Bayer's Vitraki) using NTRK fusions — also bolstered the proportion of samples with "actionable" biomarkers.

This analysis of biomarker actionability within GENIE using OncoKB captures the progress being made in precision oncology, Bedard said, not just in terms of the drug approvals but also in terms of how investigational biomarker-treatment pairs are moving up in the levels of evidence.

Following those without targetable markers

While GENIE is useful for tracking the growth in the number of patients who have biomarker-informed treatments, the repository also captures the experiences of patients who lack targetable tumor alterations. Pugh reported that in the current dataset, on average, 20 percent of tumors have no identified driver mutation. "This may be the cohort with potential for whole-genome or transcriptome comprehensive profiling," he said.

GENIE is also growing in terms of the amount of phenotypic data, including information on patients' tumor type, histology, demographics, and vital statistics. Pugh noted that investigators are eager to increase data on the medications patients are receiving and treatment outcomes.

Additionally, within a GENIE subproject, called the BioPharma Collaborative, 10 drug companies are contributing genomic and extensive phenotypic data on several thousand non-small cell lung and colorectal cancer patients. Eventually, the aim is to grow the BPC cohort to 50,000 patients with a range of tumor types. Drugmakers hope to use this data to identify novel therapeutic targets, identify new biomarkers predictive of therapy, and design better precision oncology trials.

Using the first iteration of the BPC data in NSCLC, Beilei Cai, senior director of US oncology health economics and outcomes research at Novartis, looked at the real-world progression-free and overall survival outcomes of patients who don't have EGFR mutations or ALK rearrangements. She noted that although these patients can receive immunotherapy, chemotherapy, or other targeted drugs, there is an unmet need for better treatments in this population.

The analysis included 263 advanced NSCLC patients, most of whom received NGS tumor profiling from 2014 to 2017 and were found to have EGFR or ALK wild type mutations. Cai reported that 67 percent received chemo as their first-line treatment, 15 percent received bevacizumab with chemo, 12 percent got immunotherapy, and 5 percent got a targeted drug.

Median progression-free survival according to radiologists' evaluations of imaging scans was five months for stage IV patients and by oncologists' assessment was 5.5 months, which Cai said showed that oncologists in the real-world don't just rely on imaging measurements to evaluate patients' outcomes. Overall survival was around 15 months for stage IV patients, with those on chemo having the shortest overall survival, followed by patients receiving immunotherapy, and bevacizumab-chemo; the median overall survival was not reached for the 13 patients who received targeted treatments.

As biomarker testing rates continue to increase among NSCLC patients and more drugs are approved, Cai's group hopes to repeat this analysis to track the treatments patients are receiving and their real-world outcomes on the therapies.

Understanding racial disparities

Using the GENIE data, researchers were also able to explore the differences in the molecular features of cancers in patients of different races. Andreana Holowatyj, assistant professor of medicine and cancer biology, epidemiology at Vanderbilt University, used clinical and targeted sequencing data in GENIE from 6,120 colorectal cancer patients treated at 12 participating institutions to investigate distinct molecular features of early versus late-onset disease in specific racial groups.

Although the incidence of colorectal cancer in people younger than 50 years old has been increasing over the last three decades, the complex interplay of social and biological factors that may be contributing to this, such as genetic race, sex, geographical location, and health behaviors, are poorly understood. Meanwhile, the racial disparities continue to grow, Holowatyj noted, with the incidence of early-onset colorectal cancer nearly twofold higher among non-white patients and survival "significantly worse" among Black patients compared to whites.

Holowatyj's group turned to GENIE to identify the molecular features that could underlie some of these disparities among young colorectal cancer patients of different races. A third of this cohort had early onset colorectal cancer, within which the vast majority had microsatellite stable or non-hypermutated tumors.

First, Holowatyj and colleagues confirmed what others had also seen, which is that the landscape of mutations in early-onset colorectal cancer patients is unique compared to those with late-onset disease. Early-stage tumors, for example, were more likely to have non-silent variants in TP53, LRP1B, TCF7L2, and DOCK8, and less likely to have non-silent variants in KDR and WRN.

The researchers then looked at racial differences in molecular features of early-stage tumors and found notably that non-Hispanic Black patients had a significantly higher tumor mutation burden compared to white patients. When researchers looked further into tumor molecular markers by race, they continued to find distinct genomic patterns in non-overlapping genes in early-stage tumors. For example, Asian/Pacific Islander patients had distinct mutations in PIK3CA, APC, and FAT1, while non-Hispanic Black patients had mutations in TGFRBR2 and CREBBP. These genetic mutations didn't occur significantly in white patients.  

According to Holowatyj, this is the first evidence that "molecular features of early-onset colorectal cancer may differ by race." This data, she added, may be informative for drug development and demonstrate the need to integrate mechanistic lab approaches with human studies to improve understanding of the molecular underpinnings of early-onset colorectal cancer disparities. 

Due to limited data within GENIE, Holowatyj's group wasn't able to include an analysis of Hispanic patients.

David Hein, a clinical data specialist at the University of Texas Southwestern Medical Center, also reported on the unique molecular features of early-stage colorectal cancer patients according to race. However, his team combined GENIE data with data from patients seen at his institution, which treats a lot of White Hispanic patients.

This analysis notably found that White Hispanic patients tended to have NTRK2 mutations, as well as mutations in PMS3 and other DNA repair genes. The enrichment of somatic mutations in DNA repair genes among White Hispanic patients raises the question, according to Hein, as to whether they are more likely to have a microsatellite instability-high phenotype and respond well to immunotherapy. 

Both Hein and Holowatyj noted the need to conduct further research into whether molecular differences can explain racial disparities in early-stage colorectal cancer outcomes. Holowatyj's team hopes to validate its preliminary findings, study the associations in patient-derived tumor organoids, and integrate other types of omics data. "It's vital to consider that race is a social construct," Holowatyj added, pointing out that currently GENIE lacks certain types of data, such as genetic ancestry, which limits this type of analysis.

"Our findings highlight the need for more effort in ensuring greater diversity in genetic databases so that we are adequately representing our diverse, young CRC population," Hein said.