English (United Kingdom)

https://curated-unify.zendy.io/wp-json/zendy-region/v1/featured_content/oa?rat=en

https://curated-unify.zendy.io/wp-json/zendy-region/v1/highlighted_journal/

Global: Zendy Plus annual

Presents the access of premium content as premium feature

Premium Content

Presents the keyphrase highlighting as premium feature

Keyphrase Highlighting

Presents the summarisation as premium feature

Summarisation

Insights

Presents the pdf analysis as premium feature

PDF Analysis

Presents the zaia usage as premium feature

ZAIA

Global: Zendy Plus monthly

Global: Zendy Tools annual

Global: Zendy Tools monthly

Global: Zendy Free Plan

Language models are powerful tools for molecular design. Currently, thedominant paradigm is to parse molecular graphs into linear stringrepresentations that can easily be trained on. This approach has been verysuccessful, however, it is limited to chemical structures that can becompletely represented by a graph -- like organic molecules -- while materialsand biomolecular structures like protein binding sites require a more completerepresentation that includes the relative positioning of their atoms in space.In this work, we show how language models, without any architecturemodifications, trained using next-token prediction -- can generate novel andvalid structures in three dimensions from various substantially differentdistributions of chemical structures. In particular, we demonstrate thatlanguage models trained directly on sequences derived directly from chemicalfile formats like XYZ files, Crystallographic Information files (CIFs), orProtein Data Bank files (PDBs) can directly generate molecules, crystals, andprotein binding sites in three dimensions. Furthermore, despite being trainedon chemical file sequences -- language models still achieve performancecomparable to state-of-the-art models that use graph and graph-derived stringrepresentations, as well as other domain-specific 3D generative models. Indoing so, we demonstrate that it is not necessary to use simplified molecularrepresentations to train chemical language models -- that they are powerfulgenerative models capable of directly exploring chemical space in threedimensions for very different structures.

Language models can generate molecules, materials, and protein binding sites directly in three dimensions as XYZ, CIF, and PDB files