arXiv subjects

Joshua W. Sin

Publications and source records attributed to Joshua W. Sin.

3 recordsLinked to original sources

arXiv2026

Dynamic language model representations for multi-objective reaction optimisation

Optimising chemical reactions across multiple objectives, such as yield, selectivity, and safety, is central to chemical synthesis, and model-driven approaches depend critically on how reaction components are represented. Established featurisations are either chemically uninformative, as with one-hot encodings, or, as with molecular descriptors, do not readily extend across chemically distinct components. For structurally and functionally diverse components, it is therefore unclear what a shared representation should contain. Constructing such a representation is itself a challenging research undertaking that must be revisited for each new reaction system. Here we bypass this step by learning the reaction representation dynamically from text. Textual descriptions of reaction conditions are encoded by a fine-tuned language model trained jointly with Gaussian process surrogates, yielding task-adaptive representations within a multi-objective Bayesian optimisation loop. Across nickel- and palladium-catalysed cross-couplings in both sequential and parallel experimentation regimes, this approach reaches optimisation convergence in fewer experiments than descriptor libraries or one-hot encoding. Applied prospectively to a palladium-catalysed cyanation spanning mixed ligand denticity and heterogeneous additives, and to a three-objective asymmetric hydrogenation across chiral iridium and ruthenium catalyst families, two rounds of high-throughput experimentation (192 reactions, under 3% of each design space) delivered conditions translating directly to gram scale in 94% and 84% isolated yield, the latter at 99.6% enantiomeric excess.

cs.LG↗

arXiv2026

Bi-semantic Chemical Embedder for Joint Representation Learning of SMILES and Natural Language

Transformer models have revolutionized natural language processing (NLP), and text-based molecular representations like SMILES have successfully extended these architectures to chemistry. However, domain-adaptive pre-training often causes models to overfit to chemical syntax, catastrophically forgetting their foundational semantic capabilities. To address this challenge, we introduce CheMatE, a chemistry-oriented embedding model that jointly captures molecular structure and domain-specific natural language within the same representation space. Built on a ModernBERT backbone, CheMatE learns bi-semantic representations through a two-stage training procedure: continued masked language modeling (MLM) followed by a Matryoshka contrastive learning stage via Multiple Negative Ranking Loss (MNRL). First, we train the model using MLM on a novel, large-scale corpus of SMILES-annotated, long-context scientific documents that were constructed and curated from FineWeb and ChemPile (comprising 10.4B and 11.5B tokens, respectively). Subsequently, the model undergoes contrastive learning using a synthetic dataset of SMILES-text pairs algorithmically derived from our original training corpus. This design exposes the model to SMILES-enriched scientific literature, enabling bi-semantic understanding. We evaluate CheMatE across a range of downstream tasks covering molecular property prediction and scientific language understanding. Our results demonstrate that coupling our custom-curated datasets with this sequential training strategy yields robust, highly transferable representations. By effectively unifying structural and contextual signals within a single text-based framework, CheMatE achieves competitive performance across both specialized chemistry models and general-purpose language model baselines.

cs.LG↗

arXiv2025

Swarm Intelligence for Chemical Reaction Optimisation

Chemical reaction optimisation is essential for synthetic chemistry and pharmaceutical development, demanding the extensive exploration of many reaction parameters to achieve efficient and sustainable processes. We report $α$-PSO, a novel nature-inspired metaheuristic algorithm that augments canonical particle swarm optimisation (PSO) with machine learning (ML) for parallel reaction optimisation. Unlike black-box ML approaches that obscure decision-making processes, $α$-PSO uses mechanistically clear optimisation strategies through simple, physically intuitive swarm dynamics directly connected to experimental observables, enabling practitioners to understand the components driving each optimisation decision. We establish a theoretical framework for reaction landscape analysis using local Lipschitz constants to quantify reaction space "roughness", distinguishing between smoothly varying landscapes with predictable surfaces and rough landscapes with many reactivity cliffs. This analysis guides adaptive $α$-PSO parameter selection, optimising performance for different reaction topologies. Systematic evaluation of $α$-PSO across pharmaceutically relevant reaction benchmarks demonstrates competitive performance with state-of-the-art Bayesian optimisation methods, while two prospective high-throughput experimentation (HTE) campaigns showed that $α$-PSO identified optimal reaction conditions more rapidly than Bayesian optimisation. $α$-PSO combines the predictive capability of advanced black-box ML methods with interpretable metaheuristic procedures, offering chemists an effective framework for parallel reaction optimisation that maintains methodological clarity while achieving highly performant experimental outcomes. Alongside our open-source $α$-PSO implementation, we release $989$ new high-quality Pd-catalysed Buchwald-Hartwig and Suzuki reactions.

physics.chem-ph↗