Moving oncology from a coin-flip to a calculated decision
PRECISION IMMUNOTHERAPY
Biomine predicts a immunotherapy response before treatment begins using patient-derived microbial signals, the missing host layer in precision oncology.
THE PROBLEM
Two patients. Same tumour markers. Different outcomes
Immunotherapy can be life-changing - but for most patients, it doesn't work. Checkpoint inhibitors fail 60–80% of patients. We are currently guessing who gets them.
Clinicians are forced to treat partially informative signals as decisive ones. The tumour has been measured, quantified, labeled - yet in the room where the decision is actually made, the feeling is rarely "certainty."
"A perfect map is useless in a broken car"
- Tumour markers describe the destination; they can't tell you if the vehicle can get there.
Immunotherapy Response Reality
Responders
Non-Responders
~30%
~70%
28.9%
FDA approvals where PD-L1 was predictive
0.65
AUC for PD-L1 IHC prediction
>$100K
Cost per immunotherapy course (US)
Increased death risk per month of delay
10%
Sources: JAMA Oncology meta-analysis; FDA approval analyses (2011-2019)
THE INSIGHT
The gut is the immune system's control room.
Checkpoint inhibitors don't kill cancer cells directly - they rewire the immune system. They change signaling, exhaustion states, T-cell activity. And those things are properties of the patient, not just the tumor.
THE RESPONSE EQUASION
Tumor
Biology
(Known)
Host
Readiness
We measure this
Biomine
Response
Prediction
(Calculated)
- Tumour markers describe the destination; they can't tell you if the vehicle can get there.
THE SOLUTION
A pre-treatment response score for checkpoint inhibitors.
Stool collected before treatment begins. Non-invasive. Fits existing clinical workflows.
1
Sample
2
Sequence
Shotgun metagenomics captures the microbial ecosystem. Proprietary AI extracts predictive signals.
2
3
Report
Response probability score + risk flag. Actionable intelligence for clinicians and trial sponsors.
The Biological Click
The microbiome acts as a measurable immune control layer. It constantly trains and shapes immune signaling—inflammation, T-cell activity, metabolic state.
Multiple systematic reviews have reported associations between microbial disruption (e.g., antibiotics) around checkpoint inhibitor treatment and worse outcomes.
Host-state biomarkers describe the immune system that has to do the work. The microbiome is that system's most information-rich, measurable signature.
THE TRACTION
Proof we're not starting from zero.
Recall for non-responders
86%
Version 1 engine in early validation. Unlike PD-L1 (poor rule-out) and TMB (high variability), Biomine v1 is signal-locked on the host state.
TRAINING DATASET
baseline samples
From 10 independent studies. The largest consolidated baseline training set of its kind.
938
Commercially Active
OPERATIONAL STATUS
Already shipping microbiome products end-to-end
FDA CLASS II Path
REGULATORY TAILWIND
2026 reclassification: PMA → 510(k)
Most microbiome data is Western. This creates brittle models that break when deployed across diverse populations.
Built in South Africa, Biomine has access to underrepresented microbial diversity that Western competitors cannot replicate. This isn't just a moral opportunity—it's a scientific and commercial advantage.
Training on this underexplored biological universe creates a predictor that is inherently more robust and generalizable.
THE MOAT
An unfair advantage in data diversity.
WHY THIS MATTERS
Data Fortress
Competitors in the West cannot replicate this cohort access
Compounding Advantage
Each new cohort expands the domain of validity
Execution Engine
Not learning the plumbing—already shipping real products