The Global Alliance for Genomics and Health (GA4GH) is uniting researchers worldwide to share genomic data responsibly, set open standards, and improve oncology care. Leaders from its Cancer Community discuss the current state of cancer genomics and where the field is headed
The Global Alliance for Genomics and Health (GA4GH) seeks to build the foundation for responsible and broad sharing of genomic and health data, powered by open standards and shared approaches, under a human rights framework. Through its Cancer Community, GA4GH unites cancer initiatives across the globe to enhance oncology care through the development of standards that enable cancer genomic data sharing and collaborative knowledge exchange. Three of the five co-leads of the GA4GH Cancer Community – Zinaida Perova (EMBL’s European Bioinformatics Institute), Benjamin Haibe-Kains (University Health Network), and Bernie Pope (The University of Melbourne) discuss the current landscape of cancer genomics and how the field may evolve in the future.
Q. What are the key ways in which cancer genomics research contributes to precision medicine?
ZP: Genomics plays a key role in tailoring treatments. By identifying specific molecular alterations in patients’ tumors, clinicians can match patients with treatments that are particularly effective for those characteristics. For example, the presence of HER2 amplification in breast cancer predicts response to drugs targeting HER2, and the mutation of EGFR in lung cancer predicts response to EGFR-targeted therapies.
Knowing specific genomic features is valuable, but it doesn’t always guarantee that a treatment will be effective, as cancer is a highly heterogeneous disease and tumors with the same mutations can have different responses to the same drugs. Functional precision oncology is an approach that combines static genomic features of cancers with functional testing – measuring how patients’ own tumor cells respond to drugs. This is done using patient-derived models, sometimes referred to as patient avatars, and allows for testing multiple drugs or drug combinations to guide treatment selection.
BP: Instead of applying the same treatment regimen to everyone with a specific cancer type, as may have been done historically, we recognize that even within certain cancer types, there are various subtypes. Some subtypes may have a more aggressive trajectory and develop metastatic disease faster than others.
Genomics research allows us to predict the progression of a cancer from an earlier stage. By detecting cancer at an earlier stage and sequencing the tumor’s genome, we can classify it into different risk categories.
BHK: Understanding the genomic vulnerabilities of a patient’s tumor is essential, but it is only part of the picture. Not all patients respond equally to anticancer therapies, and identifying predictive biomarkers, whether from genomic profiles or other data modalities, is critical to pinpoint those most likely to benefit, whether in standard-of-care settings or clinical trials.
Biomarker discovery hinges not only on high-quality molecular and imaging data, such as sequencing from blood or tumor biopsies, radiological scans, or pathology slides, but also on how accurately we assess therapy response. This is a frequently overlooked challenge, especially in patients with multiple metastases.
Current response evaluation methods are overly simplistic; due to logistical constraints, we often monitor only a handful of lesions, which may not reflect the full spectrum of disease progression or response, given the heterogeneity across lesions.
Q. What challenges arise from analyzing data in isolation, and how crucial is international, multi-disciplinary collaboration for advancing cancer genomics research?
BP: I’m involved in the Pan-Prostate Cancer Group (PPCG), the largest international collaboration focused on whole genome sequencing for prostate cancer, including both primary and metastatic disease. One of our goals is to identify which patients need intensive treatment early, while others may benefit from a watchful waiting approach to avoid overtreatment and its associated morbidities.
To obtain statistically significant results, we require a large dataset. The PPCG brings together global collaborators to harmonize and analyze data, ensuring that we capture the broad range of scenarios in prostate cancer. Without this integrated approach, we would lack the statistical power necessary for meaningful research outcomes.
BHK: We face several obstacles in accessing enough diverse data for predictive models in research. While our lab’s models perform well on curated datasets, they often don’t reflect real-world conditions, which include more missing data and noise. Additionally, patient populations in practice are more diverse than those in research studies, leading to biases in AI models. Sharing data at various levels – local, national, and international – is crucial to understanding the true diversity and performance of these models. As we develop more models for clinical use, this need will grow. We also want to automate patient assignment to trials, which is complex and hindered by a lack of data sharing. Some critical data, like trial protocols, are not widely shared due to the proprietary nature tied to pharma sponsors, preventing transparency and innovation in precision medicine.
ZP: Diversity in research is important, and highlights the need for diverse data sets, as findings from one ethnic group may not be applicable to others. International collaborations and data sharing efforts are crucial to drive progress in precision medicine, especially in rare cancer types and pediatric cancers, where the sample sizes are very small. Generation and analysis of data in isolation creates data standardization and harmonization issues when trying to integrate findings from different projects or institutions. GA4GH addresses these challenges by bringing together multi-disciplinary experts from all over the world to collaboratively develop standards and tools for responsible and ethical data sharing.
Q. What are some of the unknowns or gaps in cancer genomics that still need to be explored to fully understand the pathogenesis of diverse tumor types?
BHK: While research can delve deeply into various biological layers – many of which were previously unknown – there’s still a long way to go before we have a complete understanding. On the clinical side, particularly regarding precision medicine, the gap between what can be researched and what can be implemented in practice is enormous.
As many others, I began working with RNA technologies, such as gene expression microarrays, back in the 1990s. Despite the maturation of the technology and the intense research around this new data modality, only a few biomarkers or predictive models have been approved for clinical use. Although we’ve now adopted RNA sequencing in most of our trials in Canada, I still don’t think it’s commonplace in other countries. There are so many layers, including proteomics, which may not make it to the clinic for decades. This gap is significant and, in many ways, structural; we can perform many clinical procedures that are complex and costly, but we often cannot afford them.
Instead of relying on static snapshots, we should track patients over time. Liquid biopsy, which enables the detection of circulating tumor DNA, opens new avenues for early detection and monitoring. Currently, we lack a holistic view; we often focus on one biopsy without considering others or other factors affecting the patient. Family support, for example, is a key predictor of therapy response, yet we overlook it. Patient-reported outcomes could provide a more comprehensive perspective.
BP: One of the gaps that I see in the research and translation fields is that, while genomics is a fantastic tool – along with other ‘-omics’ technologies like proteomics and metabolomics – cell biology is incredibly complex and dynamic. It is heavily influenced by both the environment and the microenvironment surrounding the cells.
When we sequence a genome and extract DNA to look for somatic mutations in tumors, the information we obtain is valuable, but it represents only a small part of what is actually happening within the cell. Furthermore, this analysis provides just a static snapshot at a single point in time for a collection of cells.
I believe a significant knowledge gap exists in our understanding of the dynamic behavior of cellular systems as a whole. Although we still have a long way to go, progress is being made. Multiomics is contributing to this by enabling us to combine data from genomics, proteomics, metabolomics, lipidomics, and transcriptomics to gain a better understanding of dynamic cellular behaviors. Additionally, single-cell sequencing is helping us uncover the complexity of the cellular microenvironment. While it is fantastic that we have access to genomic information, this remains a critical gap in our current knowledge.
ZP: Tumor heterogeneity is one of the most challenging unknowns, as different regions within a tumor have different genetic profiles. In addition, the tumor continues to evolve over time. Current genomic approaches often analyze single time points or limited tissue samples, missing the complex spatial and temporal dynamics of tumor evolution. Patient-derived cancer models, generated by growing a patient’s tumor cells in a 3D culture or implanting them in a model organism, provide a useful tool to study tumor evolution and the tumor microenvironment. Models of different cancer types generated by academic labs and commercial entities are often shared through online platforms, such as CancerModels.Org, to enable reuse of associated genomic, clinical, and functional data by the research community.
One critical issue is the lack of training in genomics for medical professionals. While larger cancer centers may have more resources to conduct specific genetic screenings and interpret the results, a training gap remains that needs to be addressed. As new data types and techniques continue to emerge, ongoing education is essential for clinicians to effectively integrate and interpret genomic information. Moreover, genomic test results will remain of little value if they are not easily accessible in the patient’s medical record.
Q. How could mapping the process of early tumor development help to detect and prevent cancer at an earlier stage?
BHK: That’s a highly active area of research. I believe liquid biopsy, which enables the detection of circulating tumor DNA in the blood, has significantly advanced this field because, in a minimally invasive way, we can now track patients over time. Previously, we relied on radiological imaging and symptom reporting, which can be cumbersome and often required visits to large centers or clinics.
Now, people are considering using technologies that could even be utilized at home to provide a much more detailed follow-up of patients. For instance, Li-Fraumeni Syndrome patients are at a higher risk of developing multiple cancer types throughout their lifetime. It’s crucial for them to be monitored very closely. If we are late in detecting a tumor, it could be fatal, but if we catch it early, we can intervene and potentially extend their lives for decades.
BP: One observation emerging from recent research is that in some tumors, the precursor events or mutations that give rise to tumor lineages can occur many years, even decades, before a detectable tumor forms. This means there are latent potential tumors present long before they are identified. Currently, tumors are typically discovered through screening programs – often due to hereditary risk factors or when a patient presents with symptoms, because the tumor has already grown significantly. Late-stage detection is concerning because it usually results in worse outcomes for the individual.
On the other hand, early detection is beneficial if there are effective treatments available that do not negatively impact the patient. For example, in the case of prostate cancer, early detection can be advantageous; however, many prostate cancers tend to have a relatively indolent course, and only a small subset is aggressive and high-risk.
With early detection, oncologists face the question of what action to take. Should the patient undergo an operation that carries potential side effects? Are there alternative treatment options, or should the patient simply be monitored? Additionally, patients might feel anxious about having knowledge of a potentially risky tumor and grapple with what the best course of action is. Many may lean towards caution and consider surgery, believing it might be a prudent choice, but such decisions could lead to broader negative consequences for the population as a whole.
Thus, while early detection is important and valuable, it must be paired with a better understanding of the individual risk associated with that specific cancer.
This brings us back to the concept of precision medicine. Circulating tumor DNA is a promising diagnostic tool for many cancers. It is particularly advantageous because it is minimally invasive; a blood sample is one of the least invasive methods available. It’s reasonably non-invasive and has the potential to detect early-stage disease. But it also has the potential to detect disease recurrence after treatment, which I think is also very important because, if you’ve had a prostatectomy, for example, then there’s another issue for the individual, which is a recurrence of the disease, where the tumor grows back at the site where the prostate was removed. And that’s typically monitored through, for example, PSA levels. But that actually takes quite a while for that to develop up to a detectable level again. Whereas, at least theoretically, with circulating tumor DNA, you could detect circulating tumor DNA very quickly after treatment and see whether it is effective or whether there’s recurrence.
ZP: Over the years, we have significantly improved our ability to treat established cancers. However, if we leverage the advancements made in cancer research to focus more on prediction, such as identifying recurring evolutionary signatures or trajectories across patient cohorts, we could potentially catch cancers at earlier stages. Additionally, developing a framework that assesses individual risk for those without family history of cancer who do not qualify for routine cancer screening could be beneficial.
Combined with the data about the patient’s immune system, microbiome, and environmental exposures, this approach aligns with the holistic strategies that have been discussed, aiming to shift the focus toward prevention by utilizing the wealth of knowledge we have gained, particularly from genomics, to provide insights that can help identify risks for individuals.
Achieving this goal would be incredibly valuable. However, it does depend on the availability of comprehensive data and our ability to combine information from various sources to derive meaningful insights.
Q. How do you think the field of cancer genomics will evolve in the future?
ZP: Integration of multiple data types will provide a holistic view of the disease. AI promises to aid molecular tumor boards and bridge the gap between complex data and clinical decision-making. Proactive screening and multi-omics profiling offer the possibility of early detection and early intervention, eventually moving from precision oncology to precision cancer prevention.
Additionally, there is an opportunity to better educate patients about their health. Patient advocacy organizations already play a crucial role, and we can further enhance access to existing knowledge and build platforms that help individuals understand their genomic profiles, treatment options, and clinical trial opportunities. The future of cancer genomics will not be so much about generating more data but about making it meaningful and actionable for clinicians and patients.
BP: Cost is a critical issue in research and clinical care, especially in genomics, where assays can be expensive. While costly tests may be justified in research, clinical care requires more manageable costs. Genomic testing can be complex, involving time, equipment, and expertise, which complicates routine clinical practice.
Standardization poses additional challenges, as ideal lab conditions often differ from real-world clinical situations. Tissue samples are typically limited in quality and quantity, impacting test accuracy.
While genomics provides extensive information about tumors, such as mutations and signatures, the challenge lies in translating this data into meaningful improvements in clinical practice. If the changes are minor, the inertia in updating standards and training can prevent significant advancements. Therefore, demonstrating a substantial benefit over the current standard of care is essential for adoption.
BHK: Translating discoveries and deploying new technologies into clinical practice represent immense challenges. Our current assessment of standard care, particularly in fields like radiology, is often flawed.
We set unrealistically high expectations for algorithms while failing to recognize the shortcomings of human practices. The bottom line is that successful translation requires clinical champions; clinicians should lead and take credit for these initiatives.
When benefits are subtle, our ability to quantify standard care becomes crucial. As information access increases, particularly through molecular tumor boards, we risk overwhelming clinicians. Innovative tools, like agentic AI, show promise for improving decision-making, but there are significant obstacles to adoption.
To bridge these gaps, hospitals need dedicated teams focused on implementation science, empowering the right leaders – whether clinicians, administrators, or nurses – to facilitate the integration of new technologies into clinical settings.
