Back to ChemrytIQ Help ChemrytIQ Child Module

ChemrytIQ-DeNovo - DeNovo Analog Generation

AI-assisted medicinal chemistry design, analog generation, scoring, and prioritization from a parent molecule.

Prerequisite: Load The Molecule From ChemrytIQ

Before opening ChemrytIQ-DeNovo, search the molecule in ChemrytIQ by SMILES, InChI, molecule name, or CAS number. Confirm the correct molecule on the ChemrytIQ page, then open the required Chemryt app from that same molecule context so the selected structure is loaded into the app automatically.

What It Does

ChemrytIQ-DeNovo generates and prioritizes analog ideas from a parent molecule using medicinal chemistry rules, QSAR scores, toxicity scores, pharmacophore features, synthesis-feasibility cues, SELFIES, and future ChemrytRX reaction intelligence.

Parent molecule

Uses the active ChemrytIQ molecule as the design seed and displays structure, SMILES, SELFIES, and source state.

Generation stages

Applies rule-based transformations, analog priority models, scaffold/R-group summaries, and optional SELFIES/LSTM generation.

Prioritization

Ranks analogs by property deltas, toxicity movement, QSAR fit, pharmacophore retention, Pareto balance, and synthetic plausibility.

Quick Tutorial

  1. Load or search a parent molecule in ChemrytIQ, then open the DeNovo panel.
  2. Confirm the parent structure, identifiers, and generation capabilities before running analog generation.
  3. Choose generation options and run the analog workflow.
  4. Review generated analog cards, filters, alert groups, score deltas, scaffold chips, and SAR table rows.
  5. Compare parent versus analog descriptors, pharmacophore behavior, toxicity movement, and QSAR movement.
  6. Select promising analogs for PCP, QSAR, ToxPred, docking, or synthesis feasibility review.

Main Areas

AreaWhat to reviewWhen to use it
Pre-run panel Input source, current molecule structure, capabilities, and generation stages. Use to verify the parent and run settings.
Analog results Analog identity, transform, filters, priority score, deltas, SAR table, and copy controls. Use to shortlist generated molecules.
Decision support Pareto badge, parent-vs-analog comparison, alerts, synthesis cues, and downstream module actions. Use to decide what to explore next.

ML Model / Computation Used

Model or methodWhat it predictsImplementation details
DeNovo analog-priority Random Forest Classifies parent/analog pairs as improved or not improved for prioritization. Uses sklearn_random_forest_classifier with 500 estimators, Morgan fingerprints, BindingDB-derived analog pairs, about 80,918 training rows, and 250,374 training pairs. Artifacts are available as joblib and ONNX.
ChemrytIQ-DeNovo LSTM v1 Generates SELFIES-based de novo molecule strings. Decoder-only PyTorch LSTM language model with 2 layers, hidden size 256, 1.11M parameters, SELFIES vocabulary size 489, trained on public ChEMBL/GuacaMol/MOSES/PubChem-derived corpus shards; ONNX export supports autoregressive generation.

Good Practice

Generated molecules are hypotheses. Confirm validity, synthetic accessibility, IP position, assay relevance, and safety before using DeNovo suggestions in project decisions.

Reference Used

This Tutorial page mirrors the ChemrytIQ reference module: ChemrytIQ-DeNovo.

ChemrytIQ child-module tutorial documentation. Use computational outputs as decision support and validate important conclusions.