A large-scale benchmark of 25 molecular optimization methods on 23 oracles under constrained oracle budgets, showing that sample efficiency is a critical and often neglected dimension of evaluation.
A comprehensive survey classifying molecular representation learning foundation models by input modality (sequence, graph, 3D, image, multimodal) and analyzing four pretraining paradigms for drug discovery tasks.
Introduces SPECTRA, a framework that generates spectral performance curves to measure how ML model accuracy degrades as train-test overlap decreases across molecular sequencing tasks.
STONED introduces simple string manipulation algorithms on SELFIES for molecular design, achieving competitive results with deep generative models while requiring no training data or GPU resources.
Introduces TamGen, a target-aware molecular generation method using a pre-trained GPT-like chemical language model with protein structure conditioning. A Design-Refine-Test pipeline discovers 14 inhibitors against tuberculosis ClpP protease, with IC50 values as low as 1.9 uM.
Frey et al. discover empirical power-law scaling relations for both chemical language models (ChemGPT, up to 1B parameters) and equivariant GNN interatomic potentials, finding that neither domain has saturated with respect to model size, data, or compute.
This paper introduces ROGI-XD, a reformulation of the ROuGhness Index that enables fair comparison of QSPR surface roughness across molecular representations of different dimensionalities. Evaluating VAE, GIN, ChemBERTa, and ChemGPT representations, the authors show that pretrained chemical models do not produce smoother structure-property landscapes than simple molecular fingerprints or descriptors.
This position paper demonstrates that genetic algorithms (GAs) perform surprisingly well on molecular generation benchmarks, often outperforming complex deep learning methods. The authors propose the GA criterion: new molecule generation algorithms should demonstrate a clear advantage over GAs.
MolGenSurvey systematically reviews ML models for molecule design, organizing the field by molecular representation (1D/2D/3D), generative method (deep generative models vs. combinatorial optimization), and task type (8 distinct generation/optimization tasks). It catalogs over 100 methods, unifies task definitions via input/output/goal taxonomy, and identifies key challenges including out-of-distribution generation, oracle costs, and lack of unified benchmarks.
Proposes a benchmark for de novo drug design using SMINA docking scores across eight drug targets, revealing that popular generative models fail to outperform random ZINC subsets.
Tartarus introduces a modular suite of realistic molecular design benchmarks grounded in computational chemistry simulations. Benchmarking eight generative models reveals that no single algorithm dominates all tasks, and simple genetic algorithms often outperform deep generative models.
This paper couples a retrosynthesis transformer with a forward reaction transformer through parameter sharing, cycle consistency checks, and multinomial latent variables. The combined approach reduces top-1 SMILES invalidity to 0.1% on USPTO-50K, improves top-10 accuracy to 78.5%, and achieves 87.3% pathway coverage on a multi-pathway in-house dataset.