onkommon.com
Programme R3
Computational Pathology & Spatial AI
Can a scanned slide predict who needs expensive sequencing first?
Coming soon
The problem
Comprehensive molecular profiling is out of financial reach for most Indian cancer patients. Sequencing everybody is not a plan; deciding who benefits most is.
Pathology capacity is also a binding constraint. There are not enough onco-pathologists in India to read what is being produced, and that gap widens every year.
The questions
- Can molecular status be predicted from a scanned histology slide accurately enough to triage who should be sequenced first?
- Does the spatial organisation of a tumour and its immune environment predict immunotherapy response?
- Can AI-assisted reading extend scarce pathology capacity without degrading it?
What we are contributing
This programme is aimed squarely at cost. If a slide that has already been made can indicate who most needs a test that costs many times more, precision oncology becomes affordable for a much larger group of Indian patients.
That is the most commercially and socially valuable output on this list, and it is a question a well-funded Western group has less reason to ask.
How the work is done
- Convolutional and transformer architectures on whole-slide images
- Weakly-supervised and multiple-instance learning
- Spatial statistics and registration of molecular and image data
- Interpretability and failure-mode analysis, knowing when a model is wrong matters more than average accuracy
What it runs on
- Digitised pathology from partner laboratories under research agreement
- Public repositories
- Matched molecular data via OnKommon
What it produces
- Image-to-molecular triage models
- Spatial signature discovery
- Research-use AI reading tools
- A pathology-AI asset for the partner network
Why it is not just science
How this changes care here
- Directly reduces the cost of precision oncology for Indian patients
- Underwrites the pathology second-opinion product
- Extends onco-pathology capacity across the partner network
Funding fit
AI-in-health grants; diagnostic industry collaboration; academic partnership for slide access.
What we are looking for
- Laboratories willing to share digitised slides under research agreement
- Machine-learning researchers in digital pathology
- Funding for scanning and storage infrastructure
Deliberately empty
Publications, funding and results
Nothing appears in this section until it exists in writing. No paper in preparation, no grant under review, no result not yet published. That rule applies with particular force to a research page.