ORPHERA ORPHARMA
Research AI agent In development
Connect research evidence to explore candidates and plan studies.
OUR PURPOSE
Why? Drug repurposing
For rare diseases with small patient populations, high costs and long development timelines create barriers to new treatments. We focus on drug repurposing: exploring new therapeutic possibilities for existing approved medicines.
Through gene-expression analysis and explainable mechanism assessment, ORPHERA provides candidates and supporting evidence to help researchers focus on follow-up validation.
~10,000Known rare diseases
95%Rare diseases without an FDA-approved treatment
ORPHERA · SERVICES
From research to implementation
ORPHERA Intelligence
Research & drug-candidate intelligence
Watch tracks research developments, Compass analyzes regulatory pathways, and Signal provides drug-candidate intelligence.
ORPHERA Foundry
Custom AI development & fine-tuning
Fine-tune custom AI models using your organization’s data and workflow criteria. We support dataset preparation and review, performance validation on separate evaluation data, and on-premises installation.
RESEARCH & DEVELOPMENT
Technology and evaluation results
TECHNOLOGY 01
CREST · Gene-expression-based candidate discovery
CREST compares disease and drug gene-expression patterns to prioritize candidates for further research.
Retrospective screening placed drugs studied in preclinical and clinical research within the top 1%.
4th
of 1,170 candidatesTop 0.34%
Sirolimus × Progeria (HGPS)
10th
of 1,170 candidatesTop 0.85%
Spironolactone × Duchenne muscular dystrophy (DMD)
These retrospective rediscovery cases connect computational rankings with prior research, suggesting potential for prioritizing new drug-repurposing candidates.
Internal retrospective comparison with prior research; not validation of predictive performance on new candidates or of clinical efficacy.
TECHNOLOGY 02
ORPHARMA · Explainable mechanism assessment model
ORPHARMA 27B · Four-class agreement (acc4)
ORPHARMAGPT-5.6-sol
Standard test
ORPHARMA · After trainingSame distribution
76.5%
Before training 36.1%GPT-5.6-sol
58.0%Gene-split test
ORPHARMA · After trainingGenes held out of fine-tuning
69.9%
Before training 41.2%GPT-5.6-sol
67.6%Drug-split test
ORPHARMA · After trainingDrugs held out of fine-tuning
69.9%
Before training 39.8%GPT-5.6-sol
63.9%Gene & drug-split test
ORPHARMA · After trainingGenes & drugs held out of fine-tuning
65.2%
Before training 30.4%GPT-5.6-sol
63.0%Absolute scores · All bars use a 0–100% scale
Original scores & sample sizes
| Condition | ORPHARMA before | GPT-5.6-sol | ORPHARMA after |
|---|---|---|---|
| Standard testn = 119 | 36.1% | 58.0% | 76.5% |
| Gene-split testn = 296 | 41.2% | 67.6% | 69.9% |
| Drug-split testn = 249 | 39.8% | 63.9% | 69.9% |
| Gene & drug-split testn = 46 | 30.4% | 63.0% | 65.2% |