Generative Models for Molecule Design — Hands-on

Learn to build and deploy generative models for molecular design using RNNs, VAEs, and Diffusion. This hands-on module covers dataset curation/tokenization, model training and sampling, goal-directed optimization with property predictors, and synthesis-aware triage (rules, SA/SCScore, basic retrosynthesis stubs) . Deliverables include a reproducible pipeline that generates, filters, and prioritizes novel compounds.

Generative Models for Molecule Design (RNN/VAE/Diffusion)

Help Desk · WhatsApp

Session Index

Session 1 — Data, Tokens & Foundations Session 2 — RNN/VAE/Diffusion: Training & Sampling Session 3 — Goal-Directed & Multi-Objective Design Session 4 — Mini Capstone: Design → Filter → Prioritize

Session 1

Fee: Rs 20800

Data, Tokens & Foundations

Design datasets & curation

standardization (salts/tautomers) dedup/scaffold splits train/val/test hygiene

Representations & tokens

SMILES/SELFIES graphs (optional) validity/uniqueness/novelty

Pipelines & infra

config/seed mgmt checkpointing reproducible runs

Session 2

Fee: Rs 23800

RNN/VAE/Diffusion: Training & Sampling

Model architectures & losses

RNN (LSTM/GRU) LM VAE (KL, β-VAE) Diffusion (noise sched.)

Training & sampling controls

teacher forcing/temp/top-k validity/uniqueness checks diversity/novelty metrics

Conditioning hooks (optional)

property-conditional inputs scaffold/fragment control prompted generation

Session 3

Fee: Rs 26800

Goal-Directed & Multi-Objective Design

Property prediction in the loop

QSAR surrogates ADMET panels docking-in-the-loop (lite)

Optimization strategies

reinforcement learning (policy) Bayesian/iterative selection Pareto & weighted sums

Constraints & controls

PAINS/Lipinski/Veber scaffold retention/avoid novelty vs similarity

Session 4

Fee: Rs 30800

Mini Capstone: Design → Filter → Prioritize

Generate a targeted library & screen

Theory + Practical

Synthesis-aware triage

SA/SCScore fragment/BB coverage basic retrosynthesis stubs

Deliverables

generated set + validity/novelty ranked list (multi-objective) report + configs/notebooks

PDF

Generative Models for Molecule Design (RNN/VAE/Diffusion) | Goal-Directed & Synthesis-Aware

Generative Models for Molecule Design — Hands-on