You're pledging to donate if the project hits its minimum goal and gets approved. If not, your funds will be returned.
Cancer research produces large amounts of molecular and genetic data, but understanding how these measurements interact across biological pathways remains difficult. Existing tools often present either complex pathway diagrams or statistical predictions, leaving researchers without a clear connection between an AI model's output and the biological processes behind it.
We propose an open-source platform that combines artificial intelligence with interactive cancer-pathway visualization. It aims to help doctors and researchers develop a clearer, patient-specific understanding of how cancer may be developing. The platform will analyse each patient's molecular data in the context of known biological pathways, creating an individualized representation of the genes, proteins, mutations, and cellular processes involved in that person's cancer.
Cancer is not identical across patients, even when they share the same general diagnosis. OncoPathAI aims to make these differences visible. Instead of presenting AI predictions as a black box or displaying the same generic pathway diagram for everyone, the tool will generate a pathway view shaped by the individual patient's data. It will bring regression-based model outputs together with biological indicators such as gene and protein overexpression or underexpression, mutations, oncogene and tumour-suppressor activity, and pathway activation or inhibition. Doctors and researchers will be able to explore which pathways appear altered, how they interact, and which molecular processes may deserve closer investigation.
The first version will use public, non-sensitive cancer datasets. It will be developed for research, biological interpretation, and hypothesis generation. Clinical use would require further validation before the tool could support diagnosis or treatment decisions.
The project builds on an academic visualization prototype developed at Aristotle University of Thessaloniki. The prototype established the initial approach for displaying oncogenes, tumour-suppressor genes, expression changes, mutations, cellular compartments, and connected pathways. Funding will support the new work required to add the AI component, evaluate it, and turn the prototype into a documented and publicly accessible research platform.
The main outputs will be:
- an open-source pathway-aware AI model;
- an interactive platform for exploring AI results and biological indicators through cancer pathways;
- a public demonstration using non-sensitive research data;
- an evaluation comparing the system with simpler prediction methods;
- accessible visualizations suitable for users with colour-vision deficiencies;
- documentation that allows other researchers to understand and reproduce the work; and
- a public research report describing the results and limitations.
Every patient's cancer has a different molecular profile. These differences can affect which biological pathways are disrupted and how the disease develops. However, the scale and complexity of molecular data make these patient-specific patterns difficult to interpret.
AI models can find patterns in complex cancer datasets, but their results are difficult to trust or investigate when the reasoning remains hidden. OncoPathAI will make those results more transparent by showing how they relate to established molecular interactions and to the profile of an individual patient.
The platform aims to give doctors and researchers a more complete view of each case, helping them inspect AI-generated findings, identify potentially relevant pathways, compare patient profiles, and generate better-informed research hypotheses. Its value will come from connecting patient-specific data, biological context, prediction, and visual explanation in one open platform.
If successful, the project could provide a foundation for analysing additional cancer types and, later, other diseases involving complex molecular pathways.
Minimum budget
- $6,000: Research and software-development time
- $2,000: Computing, storage, and hosting
- $1,000: Review by biomedical researchers
- $1,000: Independent reproducibility and software review
Ideal budget - additional
- $2,000: Further development and validation of the AI and visualization platform
-$1,000: Additional review by biomedical researchers
-$1,000: Usability and accessibility testing with intended users
-$1,000: Documentation, dissemination, and maintenance of the public platform
1) Giannis Agathos
2) Alexandros Karampikas
The AI model may not outperform simpler methods, or its pathway explanations may not be sufficiently stable. These outcomes would still be informative. We will publish the comparisons and limitations and retain the open visualization and evaluation tools for future research.
Development may also take longer than expected. In that case, we will prioritize a complete and reproducible release for one cancer type rather than spreading the work across several incomplete examples.
This project has received no other funding.