STAR-VAE: A Scalable Latent-Variable Transformer for Controllable Molecular Generation
cs.LG, cs.AI, q-bio.BM
Submitted: 2025-11-04
Updated: 2026-09-22
Comments: 46 pages, 4 figures, 10 tables, and Supporting Information
Code: https://github.com/BiomedSciAI/STAR-VAE
License: http://creativecommons.org/licenses/by-sa/4.0/
The gist: Many molecular Transformers lack probabilistic latent variables for posterior inference and latent interpolation.
Terminology
Abstract
Many molecular Transformers lack probabilistic latent variables for posterior inference and latent interpolation. We introduce STAR-VAE, a SELFIES-encoded, Transformer-based, AutoRegressive Variational AutoEncoder combining a bidirectional encoder with an autoregressive decoder pretrained on 79 million PubChem molecules. A property signal jointly conditions the prior, posterior, and decoder, while LoRA adapters support fine-tuning on small datasets without modifying the backbone. STAR-VAE achieves 100% validity and near-perfect novelty under unconditional MOSES sampling, the lowest KL divergence on five of ten GuacaMol descriptors, Spearman = 0.62 at 98% validity for synthetic-accessibility conditioning, and directional docking-score control for three Tartarus protein targets. Across four ChEMBL targets, seed-based posterior sampling recovers target-associated held-out scaffolds while label-conditioned sampling produces structurally diverse outputs. Code is available at https://github.com/BiomedSciAI/STAR-VAE.
Sources
- Syntax-Directed Variational Autoencoder for Structured Data
- Domain-Agnostic Molecular Generation with Chemical Feedback
- SELF-BART : A Transformer-based Molecular Representation Model using SELFIES
- Multi-view biomedical foundation models for molecule-target and property prediction
- Self-Referencing Embedded Strings (SELFIES): A 100% robust molecular string representation
- Adversarial Autoencoders
Related papers
- Polynomial-Augmented Neural Networks (PANNs) with Weak Orthogonality Constraints for Enhanced Function and PDE Approximation
- AIRL-S: Unifying Reinforcement Learning and Search-Based Test-Time Scaling via Adversarial Inverse Reinforcement Learning
- Transformers as Bayesian In-Context Experimenters: Smoothness-Adaptive Efficient ATE Estimation
- Convergence issues in Relational Concept Analysis based on AOC-posets
- Beliefs Beyond Posteriors: Local-Consistency Optimisation for Bayesian Neural Networks
- Understanding Diffusion Models via Ratio-Based Function Approximation with SignReLU Networks