01
Diagnosis waits for surgery
Molecular results arrive around three weeks after surgery. Frail patients and deep tumours may never get one.
Brain tumour liquid biopsy
Reading the brain through blood.
Cerelume is developing a blood-based test that detects glioma and identifies its molecular class through cell-free DNA methylation, enabling earlier diagnosis, ongoing monitoring and treatment decisions without waiting for surgery.
Research-stage platform · Not yet commercially available
Atlas data provenance
74 datasets · 8 countries · 25,750 methylation arrays
Hover or tap a country for its centres
Methylation arrays per country
- under 100
- 100 to 999
- 1,000 to 9,999
- 10,000 or more
- no dataset
- centre city
15,606
harmonised methylation profiles in the atlas
99.2%
glioma subtype accuracy, leakage-controlled
0.90 to 0.96
AUC in an independent 68-patient cohort
<0.1%
tumour-derived fraction of cell-free DNA
Our science · The problem
Everything about a glioma depends on molecular class. Today that only comes from tissue.
Since the 2021 WHO classification, diagnosis, prognosis and treatment of glioma depend on molecular class. That information comes only from tissue obtained by surgery, and there is no blood test for glioma in routine clinical use anywhere.
02
Progression is seen late
Only re-operation can confirm transformation.
03
Resistance is found by failure
Nothing tracks it without repeat tissue.
04
Recurrence relies on an ambiguous scan
Pseudoprogression mimics recurrence in a quarter of patients.
05
Spread and mimics need tissue
Metastasis, lymphoma and leptomeningeal disease are hard to separate on imaging.
Existing liquid biopsies have not solved this.
The blood-brain barrier keeps circulating tumour DNA below 0.1 per cent of cell-free DNA, so mutation-based tests and multi-cancer early-detection tests report poor sensitivity for brain tumours. Cerebrospinal fluid tests perform better but need a lumbar puncture and cannot be used for frequent monitoring.
Our science · The solution
A blood test that reports glioma and its molecular class.
Cerelume reads cell-free DNA methylation with MethyGlio, our interpretable deep learning classifier. Methylation is a broad, information-rich signal that can reveal tumour identity even when the tumour fraction is faint.
The largest atlas built for this purpose
MethyGlio is trained on 15,606 harmonised methylation profiles: 6,750 CNS tumours, 7,653 non-CNS tumours and 1,203 normal tissues.
Classifies the way a clinician thinks
A hierarchical Transformer answers six questions in sequence: tumour or not, CNS or not, tumour entity, glioma subtype, grade and molecular class (IDH, 1p/19q, H3 and others).
Externally validated on a different platform
In an independent consecutive neurosurgical cohort of 68 patients profiled on newer array and sequencing chemistry, AUC was 0.90 to 0.96 across all classification levels with no retraining.
Explainable, and used to design the blood test
Integrated Gradients shows which CpG sites drive each call. We used those attributions to select loci that survive in blood, then confirmed serum-to-tumour concordance at single-locus resolution in matched patient samples by whole-genome bisulfite sequencing. The panel includes novel proprietary biomarkers.
The Cerelume platform
From a faint methylation pattern to an interpretable tumour profile.
The product is a targeted assay on a few dozen CpG sites, not whole-genome sequencing. MethyGlio combines atlas-scale methylation data, a hierarchical Transformer model and feature-level interpretation in one pathway.
01
Blood sample
Cell-free DNA carries sparse tumour-specific methylation fragments, typically below 0.1 per cent of the total.
02
Targeted assay
A panel of a few dozen CpG sites, chosen because the model relies on them and they survive in serum.
03
Hierarchical read
MethyGlio resolves tumour status, CNS origin, entity, glioma subtype, grade and molecular class in sequence.
04
Interpretable result
Glioma present or absent, likely subtype, grade signal, IDH and 1p/19q status, a confidence score and the sites behind the call.
Intended uses, in order of launch
Each use answers a question that today requires tissue.
First
Molecular classification when tissue is unavailable or non-diagnostic.
Second
Monitoring for progression, transformation and treatment resistance.
Third
Triage of symptomatic patients and high-risk groups, and later an add-on to multi-cancer panels through partnership.
Evidence, not a black box
Designed for scientific scrutiny.
MethyGlio was evaluated with locked, leakage-controlled testing, then tested on a cohort it had never seen, on a platform it was not trained on. The blood assay is at proof-of-concept stage and is not yet validated for clinical use.
Atlas scale
15,606 harmonised samples: 6,750 CNS tumours, 7,653 non-CNS tumours and 1,203 normal tissues, drawn from public repositories worldwide.
External validation
68 consecutive neurosurgical patients profiled on newer array and sequencing chemistry: AUC 0.90 to 0.96 across all classification levels with no retraining.
Platform independence
Validated across 450K, EPIC and EPIC v2 arrays and whole-genome bisulfite sequencing, so the model can move to a low-cost targeted assay.
Explainability
Every call comes with the CpG sites behind it, which supports clinician trust, regulatory review and rapid assay design.
Who we serve
Built around the people moving brain tumour care forward.
Clinical teams
Neuro-oncologists, neurosurgeons and neuroradiologists at cancer centres and large neuroscience centres who need molecular class without another operation, and a way to follow the tumour between scans.
Trial sponsors
Teams developing IDH inhibitors and other glioma therapies who need blood-based patient identification, molecular endpoints and companion diagnostic development.
Research collaborators
Academic and translational groups validating cfDNA methylation candidates, pathology laboratories seeking a validated tissue classifier, and multi-cancer screening programmes where brain tumours remain the gap.
Patients, in launch order
The people for whom tissue is not an option, or not an option again.
- 01
Adults with a suspected glioma where biopsy is high risk, impossible or non-diagnostic.
- 02
Patients with IDH-mutant lower-grade glioma on watch-and-wait or IDH inhibitors who need molecular monitoring for transformation and resistance.
- 03
Post-treatment glioblastoma patients where MRI is equivocal for recurrence.
Our company
Clinical insight meets scientific translation.
One founder owns the science and the clinic, the other owns trial delivery, health-system implementation and company building.
JG
Jay Gray
MD, PhD, MBA · Co-founder and Chief Executive Officer
Physician-founder with 15 years across oncology, clinical research, diagnostics and health technology. Ran oncology trials end to end, built an international medical function, and implemented an AI-enabled diagnostic pathway with the NHS in rare disease. Founded LASERLY Technology and took it to four live pilots in eight months.
Executive MBA, London Business School · Clinical Trial Management, University of Oxford
jay@cerelume.bioAM
Aierpati Maimaiti
MD, PhD · Co-founder and Chief Scientific Officer
Neurosurgeon with a PhD in Neurosurgery and Neuroscience and a guest PhD researcher at Maastricht University. Built MethyGlio end to end, from data harmonisation and the model to explainability, wet-lab validation and the prospective clinical cohort. 50 peer-reviewed papers, 32 as first author, co-first author or corresponding author, h-index 12.
Invited orals at ESMO 2025, MAP 2025 and MAP 2026 · ESMO Merit Award
aierpati@cerelume.bioCollaborate with Cerelume