Interactive hub demo
Run the full OpenAlgo flow on a seeded Hub paper
This walkthrough starts from the Hub example for Random Forest Models for ADMET Property Prediction with RDKit Descriptors and follows the same review, gap-analysis, generation, and validation sequence as a real translation session.
Descriptor classification walkthrough
Review the extracted method, inspect the reproducibility gaps, and generate a template-first Python project for a descriptor-based ADMET classifier.
Template family
fingerprint_descriptor_classification
Detected domain
supported
Confidence
91/100
Primary metric
ROC-AUC
This walkthrough runs entirely in the browser. It does not create a saved translation session or require Supabase.
Review extraction
Inspect the seeded method record and edit any field that feels too optimistic.
Generate template
Produce the scaffold, config, reproducibility gap report, and validation summary.
Inspect outputs
Browse the generated files exactly where a real translation session would expose them.
Reproducibility gap report
The gap report updates live as you edit the extraction.
Specified
Dataset
Curated ADMET assay panel
Data
Molecular features
RDKit molecular descriptors
Methods
Model
Classical ML (Random Forest, SVM, XGBoost, etc.)
Methods
Split strategy
Repeated random 80/20 split
Validation
Primary metric
ROC-AUC
Results
Random seed
Random seed mentioned in the paper
Salt stripping
Salt handling discussed in the paper
Hyperparameter tuning
Hyperparameter search strategy described
RDKit version
Exact RDKit version specified
Feature scaling
Feature scaling/normalization mentioned
Class imbalance handling
Class imbalance strategy described
Train/test ratio
Exact split ratio reported
Partially specified
Not specified
Stereoisomer handling
Not addressed — keeping stereochemistry as-is
Demo status
Demo extraction is ready for review. Confirm the fields, then generate the template artifact set.
Save review before generation so the flow mirrors the real translation state machine.
Reviewer note
Notes stay local to this demo and are not sent anywhere.