PASITHEA adapts deep dreaming from computer vision to molecular design, directly optimizing SELFIES-encoded molecules for target chemical properties via gradient-based inversion of a trained regression network.
Applies the Transformer architecture to generate drug-like molecules conditioned on protein amino acid sequences, treating target-specific de novo drug design as a sequence-to-sequence translation problem.
An extensive benchmark showing that training RNN generative models with randomized (non-canonical) SMILES strings yields more uniform, complete, and closed molecular output domains than canonical SMILES.
Overview of REINVENT 4, an open-source generative molecular design framework from AstraZeneca that unifies RNN and transformer generators within reinforcement learning, transfer learning, and curriculum learning optimization algorithms.
An early and influential review cataloging 45 papers on deep generative modeling for molecules, comparing RNN, VAE, GAN, and reinforcement learning architectures across SMILES and graph-based representations.
Compares RNN-based and Transformer-based chemical language models across three molecular generation tasks of increasing complexity, finding that RNNs excel at local features while Transformers handle large molecules better.
This paper introduces structured state space sequence (S4) models to chemical language modeling, showing they combine the strengths of LSTMs (efficient recurrent generation) and GPTs (holistic sequence learning) for de novo molecular design.
A GRU-based sequence-to-sequence model that learns fixed-length molecular fingerprints by translating SMILES strings to themselves, enabling unsupervised representation learning for drug discovery tasks.
SMI-TED introduces encoder-decoder chemical foundation models (289M parameters) pre-trained on 91 million PubChem molecules, achieving strong results across property prediction, reaction yield, and molecule generation benchmarks.
A Transformer-based encoder-decoder pre-trained on 861K SMILES from ChEMBL24 produces 1024-dimensional molecular fingerprints that outperform ECFP and graph convolutions on 5 of 10 MoleculeNet tasks in low-data settings.
Introduces Atom Pair Encoding (APE), a chemistry-aware tokenizer for SMILES and SELFIES, and shows it consistently outperforms Byte Pair Encoding in RoBERTa-based molecular property classification on BBBP, HIV, and Tox21 benchmarks.
SMILES-BERT pre-trains a Transformer encoder on 18M+ SMILES from ZINC using a masked recovery task, then fine-tunes for molecular property prediction, outperforming prior methods on three datasets.