Winter et al. propose CDDD, a translation-based encoder-decoder that learns continuous molecular descriptors by translating between equivalent chemical representations like SMILES and InChI, pretrained on 72 million compounds.
A minireview of chemical language models for de novo molecule design, covering SMILES and SELFIES representations, RNN and Transformer architectures, distribution learning, goal-directed and conditional generation, and prospective experimental validation.
A comprehensive benchmark evaluating GPT-4, GPT-3.5, Davinci-003, Llama, and Galactica on eight practical chemistry tasks, revealing that LLMs are competitive on classification and text tasks but struggle with SMILES-dependent generation.
CogMol uses a SMILES VAE and multi-attribute controlled sampling (CLaSS) to generate novel, target-specific drug molecules for unseen SARS-CoV-2 proteins without model retraining.
This paper fine-tunes GPT-3’s ada model on SMILES strings for classifying electronic properties (HOMO, LUMO) of organic semiconductor molecules, finding competitive accuracy with graph neural networks and exploring robustness through ablation studies.
This perspective from Choi et al. reviews foundation models in chemistry, categorizing them as ‘small’ (domain-specific, e.g., property prediction, MLIPs, inverse design) and ‘big’ (multi-domain, e.g., multimodal and LLM-based). It surveys pretraining strategies, key architectures (GNNs and language models), and outlines future directions for scaling, efficiency, and interpretability.
Group SELFIES extends SELFIES with group tokens representing functional groups and substructures, maintaining chemical robustness while improving distribution learning and molecular generation quality.
A foundational review surveying how deep generative models (VAEs, GANs, reinforcement learning) enable inverse molecular design, covering molecular representations, chemical space navigation, and applications from drug discovery to materials engineering.
LLM-Prop uses the encoder half of T5, fine-tuned on Robocrystallographer text descriptions, to predict crystal properties. It outperforms GNN baselines like ALIGNN on band gap and volume prediction while using fewer parameters.
Proposes Captions as Representations (CaR), where ChatGPT generates textual explanations for SMILES strings that are then used to fine-tune small language models for molecular property prediction.
Demonstrates that standard transformer language models, trained with next-token prediction on sequences from XYZ, CIF, and PDB files, can generate valid 3D molecules, crystals, and protein binding sites competitive with domain-specific 3D generative models.
MaCBench evaluates frontier vision language models across 1,153 chemistry and materials science tasks spanning data extraction, experimental execution, and data interpretation, uncovering fundamental limitations in spatial reasoning and cross-modal integration.