From a research question
to a testable discovery plan.
Bring a target, find public evidence, and ask your AI to connect the findings. Explore candidate concepts, a synthesis or development flow, and the computational work needed to test them.
1Define the research brief
Use public or non-identifiable research inputs. Nothing here is a validated drug or vaccine design.
No search yet. Evidence lookup works without an AI key.
2Review the evidence
Up to eight matching Europe PMC records and an optional PubChem compound. Matching records may be irrelevant; review before using them.
Your retrieved records will appear here, with source links and available abstracts.
3Infer, synthesize & design the next test
Your BYOK AI creates the report. Hypotheses and candidate structures remain unvalidated until independent computation and experiments support them.
AI has not generated a report.
The report will include evidence synthesis, mechanistic inference, candidate concepts, a conceptual development flow, a simulation plan and validation gates.
Calculate molecular properties
Run the bundled RDKit engine in your browser, then ask your AI to interpret the measured descriptors in context. These properties do not establish target activity or drug safety.
No model has been run.
RDKit validates a supplied structure and calculates molecular descriptors. DeepChem, REINVENT4, AiZynthFinder, DiffDock and MHCflurry require separate model workers; the AI simulation plan does not execute them.
Open-source discovery infrastructure
A curated model plan uses these projects where appropriate. Repository capabilities and code licenses were checked on 19 October 2026. Model weights, training data and dependencies can have separate terms.
For vaccine evidence, also review the Immune Epitope Database. Model predictions need assay validation, population coverage review and safety evaluation. No model can establish clinical effectiveness from this brief.
About this tool
What is Drug & vaccine discovery?
Drug and vaccine discovery turns a research brief (target or antigen, indication, modality, objective) into public evidence from Europe PMC and PubChem, optional real RDKit molecular descriptors, and a BYOK AI report with candidate concepts, a conceptual development flow, a validation plan and decision gates.
Key features
- Small-molecule, biologic and vaccine tracks
- Europe PMC literature and PubChem compound evidence
- Real RDKit descriptors computed in your browser and kept separate from AI output
- Candidate hypotheses, conceptual synthesis or development flow, simulation plan and decision gates
- Explicit provider-sharing consent and JSON study export
How it works
- Define the target, indication, modality and objective
- Retrieve public literature and compound evidence
- Optionally compute RDKit properties for a reviewed SMILES
- Consent to sharing and generate the BYOK discovery report
Frequently asked questions
Does RadioGx design real drugs or vaccines?
No. It produces research hypotheses and conceptual plans. Only the RDKit descriptors are executed computations; activity, safety, immunogenicity and synthesis feasibility need laboratory validation.
Which open-source models does Discovery reference?
RDKit runs in your browser. DeepChem, REINVENT4, AiZynthFinder, DiffDock, MHCflurry and Therapeutics Data Commons are documented extension points that need separately deployed workers and licences.