Patient Stratification
Characterize molecularly defined patient populations before your study starts to support enrichment designs
The Challenge
High Trial Failure Rates
A majority of novel therapies in hematological cancers are tested in molecularly unselected populations, contributing to high Phase II/III attrition rates.
Wasted Time and Capital
Significant time and capital invested in trials that fail due to molecularly unselected enrollment, not drug inefficacy. The drug may work in a molecular subgroup that was never identified.
Limited Biomarker Approaches
Traditional clinical markers (age, cytogenetics, LDH) offer limited stratification power. Multi-omics signatures can better characterize molecular subgroups.
Post-Hoc Subgroup Analysis Fails
Discovering biomarkers after trial completion requires expensive follow-up studies and can significantly delay development timelines.
"We ran a Phase II trial that failed overall, but a subset of patients had dramatic responses. If we had enrolled only those patients, we would be filing for approval today."
The Helomnix Solution
Helomnix uses AI-powered patient digital twin representations to characterize molecular profiles BEFORE trial enrollment. By integrating multi-omics data (RNA-seq, scRNA-seq, proteomics, exome), we create a unified molecular representation of each patient and project them onto our proprietary Digital Twin Map.
The Digital Twin Map visually clusters patients by molecular similarity. When combined with clinical outcomes from our curated biobanks (MM, AML, DLBCL), machine learning models characterize which molecular neighborhoods associate with distinct pathway activity and compound–biology relationships.
This supports enrichment study designs: characterize molecularly defined subgroups, supporting enrichment study designs with the potential for reduced sample size requirements.
Unique Differentiator
Unlike generic platforms, we specialize in hematological malignancies with proprietary multi-omics biobanks with clinical outcome annotations. Our disease-specific Digital Twin Maps are purpose-built for hematological malignancy stratification.
How It Works
Multi-Omics Integration
We integrate your trial candidates' multi-omics data (bulk RNA-seq minimum, enhanced with scRNA-seq/proteomics if available) into unified patient digital twins using our proprietary Multi-Modal Integration Engine.
Digital Twin Projection
Each patient is projected onto our disease-specific Digital Twin Map™, built from proprietary biobank cohorts with clinical outcome annotations.
AI-Powered Classification
Machine learning models trained on our biobank characterize each candidate's molecular profile based on their position in the Digital Twin Map and molecular similarity to annotated profiles in the reference biobank.
Enrichment & Enrollment
You receive a ranked list of patients with molecular similarity scores. Profiles can inform enrichment design considerations or stratification into molecular subgroup arms.
Real-World Application
A pharma partner was developing a novel compound for relapsed/refractory multiple myeloma. Unselected populations showed limited and heterogeneous compound–disease alignment, requiring large sample sizes.
Before
Traditional design: unselected R/R MM population requiring large cohorts and extended enrollment periods.
After
Helomnix characterized a molecularly defined subgroup using baseline transcriptomic data. The enriched design focused on a molecularly favorable context, with substantially improved mechanistic alignment.
Outcome
The study design supported a smaller cohort and faster enrollment. A companion biomarker signature was identified prospectively, supporting a streamlined development pathway.
Value to Your Organization
Cost Efficiency
Support reduced trial costs through smaller sample sizes and faster enrollment with molecularly informed enrichment designs.
Time Efficiency
Support faster trial readout with smaller cohorts and avoid expensive post-hoc biomarker discovery studies.
Enrichment Design
Support improved Phase II/III success rates by enabling enrollment of molecularly characterized subgroups.
Our Methodology
Data Inputs
- Bulk RNA-seq (minimum required)
- scRNA-seq (optional, enhances accuracy)
- Proteomics (optional)
- Whole exome sequencing (optional)
- Clinical metadata (age, stage, prior therapies)
AI/ML Techniques
- Multi-modal latent dimension extraction
- Topology-preserving dimensionality reduction for Digital Twin Map construction
- Ensemble and representation learning methods for classification
- Feature attribution analysis for biomarker interpretation
- Cross-validation on proprietary biobank cohorts
Deliverables
- Patient molecular similarity scores
- Ranked list for enrichment enrollment
- Digital Twin Map visualization
- Companion diagnostic biomarker signature
- Regulatory-ready statistical validation report
Discuss a Translational Application
We welcome discussions about how this approach can support your translational research.