InChI (International Chemical Identifier) is an open standard from IUPAC that represents molecular structures as hierarchical, layered strings optimized for database interoperability, unique identification, and web search via its hashed InChIKey.
A comprehensive review of how solid-state materials can be numerically represented for machine learning, spanning structural features, graph neural networks, compositional descriptors, transfer learning, and generative models for inverse design.
BioT5 uses SELFIES representations and separate tokenization to pre-train a unified T5 model across molecules, proteins, and text, achieving state-of-the-art results on 10 of 15 downstream tasks.
ChemGE uses grammatical evolution over SMILES context-free grammars to generate diverse drug-like molecules in parallel, outperforming deep learning baselines in throughput and molecular diversity.
Galactica trains a decoder-only Transformer on a curated 106B-token scientific corpus spanning papers, proteins, and molecules, achieving strong results on scientific QA, mathematical reasoning, and citation prediction.
The Grammar VAE replaces character-level decoding with context-free grammar production rules, using a stack-based masking mechanism to guarantee that all generated SMILES strings are syntactically valid. Applied to molecular optimization and symbolic regression, it learns smoother latent spaces and finds better molecules than character-level baselines.
LatentGAN decouples molecular generation from SMILES syntax by training a Wasserstein GAN on latent vectors from a pretrained heteroencoder, enabling de novo design of drug-like and target-biased compounds.
LlaSMol fine-tunes Mistral, Llama 2, and other open-source LLMs on SMolInstruct, a 3.3M-sample instruction tuning dataset covering 14 chemistry tasks. The Mistral-based model outperforms GPT-4 and Claude 3 Opus across all tasks.
Ertl et al. train a character-level LSTM on 509K bioactive ChEMBL SMILES and generate one million novel, diverse molecules whose physicochemical properties, substructure features, and predicted bioactivity closely match the training distribution.
MolFM pre-trains a multimodal encoder that fuses 2D molecular graphs, biomedical text, and knowledge graph entities through fine-grained cross-modal attention, achieving strong gains on cross-modal retrieval, molecule captioning, text-based generation, and property prediction.
MoMu pre-trains dual graph and text encoders on 15K molecule graph-text pairs using contrastive learning, enabling cross-modal retrieval, molecule captioning, zero-shot text-to-graph generation, and improved molecular property prediction.
DMP combines a SMILES Transformer and a GNN branch during pre-training, using masked language modeling plus a BYOL-inspired dual-view consistency loss to learn complementary molecular representations.