By Ghmkin Hassan, Postdoctoral Fellow from UCSF, Session Co-Lead
When asking cancer researchers about what cancer systems biology requires, there will be many points of agreement. However, when asking how to achieve this, an interesting discussion can be had. This was the basis of our two 50 minute roundtable sessions, which took place at the 2026 MC2 Center All Consortia Meeting in North Bethesda, MD, October 20 – 22nd.
Why we met
Our session was driven by the perspective paper ‘A Roadmap for the Future of Systems Biology in Cancer Research’ (Wiley et al., Cancer Research, 2025). The first action point within the paper describes that multiscale frameworks are needed to develop cancer models. Whilst this can be achieved, there are no clear descriptions of how. We aimed to begin discussing how to achieve this. We invited individuals from across the consortia, with varied backgrounds and areas of research to ensure that everyone had something to contribute.
What we did
We began with an overview of the paper, highlighting key challenges, opportunities and how they could shape the future. We then moved onto Action 1, outlining four key questions to ask and provided an operational definition of what a multiscale model is. We then asked participants to move to the table with the question they were most interested in. The four questions were centered around:
- Data Collection: What multiscale data are required? How can we standardize and integrate?
- Model Integration: How can we integrate data across scales? What perturbations should we focus on?
- Heterogeneity: What heterogeneity should we include in models? How can we simplify without reducing predictive power?
- AI/ML & Mechanistic Modelling: When is combining these beneficial for multiscale modelling? How can we combine them?
How the Session Went
There was a lot of energy in the room. Once people had moved to their tables, discussions began. Each table worked on their question, sharing ideas and considering possible solutions. What stood out was that each table had a mix of people from various backgrounds, areas of research, interests, teams and priorities. This really encouraged people to think about their assumptions, challenge assumptions from others, and view problems from a new perspective. This type of discussion rarely occurs within a single lab/programme and therefore meant that the 50 minute session felt very productive.
Findings from the tables
Data Collection
This table highlighted a simple truth: all data are important. However, standardizing integration of data across platforms is difficult. They noted that there is a lot of data available, with many data repositories and integration packages. However, there are issues with these:
- Packages are difficult to adapt to your own data.
- Not all data are available.
- Data from different sources are discrete and must be carefully integrated.
- There is no general guidance on data collection across organizations.
The table concluded that without better standardization of data integration, we will not be able to get a multiscale view of cancer and therefore solutions. A proposed next move is to bring multi-omics experts and consortia together, ideally with dedicated funding and communication frameworks.
Model Integration
This table discussed how to integrate experimental, computational, and clinical phenotypic data across different scales (length, time, number of samples, frequency of measurements). They described how this is currently happening by integrating data, using agent-based platforms and generating hypotheses from models that can be validated experimentally. For example, once a paper has been published, the corresponding PI should reach out to 4-5 labs to validate findings one scale higher. Similarly, molecular findings could be validated in tissue through collaborative partnerships. This is a simple way to collaborate and validate findings across scales.
Heterogeneity
This was the largest table, stimulating a rich discussion of many types of heterogeneity: spatial neighbourhoods, transcriptional states, extracellular vesicles, cell signaling, as well as proteomic and chromosomal heterogeneity. They also highlighted the importance of viruses in cancer. Viruses can be seen across the genome, adding extra variability post tumour initiation. Several challenges were identified, particularly the difficulty of demonstrating the direct effects of viruses in cancer and distinguishing their contributions from other sources of tumour heterogeneity. The group also raised questions around how these different factors influence cancer development and progression, and how reliably biomarkers and mutations can predict long-term outcomes.
They concluded that there is heterogeneity at many levels, due to different causes. Addressing this is multidisciplinary, necessitating a systems approach from the top down to bridge cancer biology and virology. For example, including epidemiologists and non cancer biologists in the conversation.
AI/ML & Mechanistic Modelling
This group had a very specific goal to integrate cell-cell interaction modelling with metabolomics modelling to integrate biological scales and increase understanding of cancer biology in relation to drug response. They highlighted that there are no known equations for cell-cell communication. One approach is to define embedding networks with metabolism as constraints. They mentioned protein structure prediction as a possible area to build on. To make progress they would need to understand the biology in more depth and competition modelling.
Overall Themes
When looking at each table, there are a few themes that emerge that transcend the scientific topics they discussed:
- People over pipelines: Each table mentioned that the next step would be bringing in the correct experts. The issue is connecting.
- Standardization and harmonization: This was mentioned repeatedly.
- Bridging disciplines: Viruses, mutations, epidemiology, multi-omics and AI were seen as critical to accurately model real tumours.
- Simple next steps: Flag passing was a simple next step that shows they want to get things done.
Looking ahead
The roundtable did not set out to settle how multiscale cancer models should be built, and it did not. It did show that the community agrees on the direction and is ready to work out the practical steps. Across all four tables, the same messages came up: the right people need to be connected, data integration needs better standardization, and the hardest problems will only be solved by bringing together disciplines that rarely share a room. Ideas such as flag passing show that progress doesn’t have to wait for large new infrastructure. Small, collaborative steps, like asking a neighbouring lab to validate a finding one scale up, can start building the bridges between data and modelling now. The 50 minutes were only a starting point, but they showed the value of mixing expertise, and we hope the conversations continue across the consortia, turning shared questions into joint projects.