This study applies perplexity, a model-intrinsic metric from NLP, to rank de novo molecular designs generated by SMILES-based chemical language models and introduces a delta score to detect pretraining bias in transfer-learned CLMs.
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.
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.
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.
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.
BARTSmiles pre-trains a BART-large model on 1.7 billion SMILES strings from ZINC20 and achieves the best reported results on 11 classification, regression, and generation benchmarks.
This study benchmarks RNN-based chemical language models against graph generative models on three challenging tasks: high penalized LogP distributions, multi-modal molecular distributions, and large-molecule generation from PubChem. The LSTM language models consistently outperform JTVAE and CGVAE.
LIMO combines a SELFIES-based VAE with a novel stacked property predictor architecture (decoder output as predictor input) and gradient-based reverse optimization on the latent space. It is 6-8x faster than RL baselines and 12x faster than sampling methods while generating molecules with nanomolar binding affinities, including a predicted KD of 6e-14 M against the human estrogen receptor.
This paper trains a transformer model to correct invalid SMILES produced by de novo molecular generators (RNN, VAE, GAN). The corrector fixes 60-95% of invalid outputs, and the fixed molecules are comparable in novelty and similarity to valid generator outputs. The approach also enables local chemical space exploration by introducing and correcting errors in existing molecules.
The Molecular Transformer applies the Transformer architecture to forward reaction prediction, treating it as SMILES-to-SMILES machine translation. It achieves 90.4% top-1 accuracy on USPTO_MIT, outperforms quantum-chemistry baselines on regioselectivity, and provides calibrated uncertainty scores (0.89 AUC-ROC) for ranking synthesis pathways.