Jove
Visualize
Contact Us
JoVE
x logofacebook logolinkedin logoyoutube logo
ABOUT JoVE
OverviewLeadershipBlogJoVE Help Center
AUTHORS
Publishing ProcessEditorial BoardScope & PoliciesPeer ReviewFAQSubmit
LIBRARIANS
TestimonialsSubscriptionsAccessResourcesLibrary Advisory BoardFAQ
RESEARCH
JoVE JournalMethods CollectionsJoVE Encyclopedia of ExperimentsArchive
EDUCATION
JoVE CoreJoVE BusinessJoVE Science EducationJoVE Lab ManualFaculty Resource CenterFaculty Site
Terms & Conditions of Use
Privacy Policy
Policies

Related Concept Videos

Molecular Models02:00

Molecular Models

Physical models representing molecular architectures of chemical compounds play essential roles in understanding chemistry. The use of molecular models makes it easier to visualize the structures and shapes of atoms and molecules.
Predicting Molecular Geometry02:27

Predicting Molecular Geometry

VSEPR Theory for Determination of Electron Pair Geometries

You might also read

Related Articles

Articles linked to this work by shared authors, journal, and citation graph.

Sort by
Same author

Machine-Learned Leftmost Hessian Eigenvectors for Robust Transition State Finding.

Journal of chemical theory and computation·2026
Same author

Energetics of Noncovalent Interactions of Protein-Ligand Complexes for Drug Discovery.

Journal of chemical information and modeling·2026
Same author

A unified multi-stream diffusion framework for robust video camouflaged object detection.

Neural networks : the official journal of the International Neural Network Society·2026
Same author

Sensing the acidity of hydrogen bond networks.

Physical chemistry chemical physics : PCCP·2026
Same author

Pharmacological rescue of mitochondrial dysfunction, neurite degeneration, and premature death of ALS and AD iPSC-derived neurons.

bioRxiv : the preprint server for biology·2026
Same author

LinkLlama: Enabling Large Language Model for Chemically Reasonable Linker Design.

bioRxiv : the preprint server for biology·2026

Related Experiment Video

Updated: May 13, 2026

Modeling Ligands into Maps Derived from Electron Cryomicroscopy
09:30

Modeling Ligands into Maps Derived from Electron Cryomicroscopy

Published on: July 19, 2024

SmileyLlama: modifying large language models for directed chemical space exploration.

Joseph M Cavanagh1, Kunyang Sun1, Andrew Gritsevskiy2

  • 1Kenneth S. Pitzer Theory Center and Department of Chemistry, University of California, Berkeley, Berkeley, CA, USA.

Nature Computational Science
|May 11, 2026
PubMed
Summary

Large language models (LLMs) were fine-tuned into SmileyLlama for drug discovery, generating novel molecules with desired properties. This approach enhances molecule design and can be applied to various scientific fields.

More Related Videos

Curation of Computational Chemical Libraries Demonstrated with Alpha-Amino Acids
08:21

Curation of Computational Chemical Libraries Demonstrated with Alpha-Amino Acids

Published on: April 13, 2022

Related Experiment Videos

Last Updated: May 13, 2026

Modeling Ligands into Maps Derived from Electron Cryomicroscopy
09:30

Modeling Ligands into Maps Derived from Electron Cryomicroscopy

Published on: July 19, 2024

Curation of Computational Chemical Libraries Demonstrated with Alpha-Amino Acids
08:21

Curation of Computational Chemical Libraries Demonstrated with Alpha-Amino Acids

Published on: April 13, 2022

Area of Science:

  • Artificial Intelligence
  • Computational Chemistry
  • Drug Discovery

Background:

  • Large language models (LLMs) show potential beyond natural language processing.
  • Exploring chemical space for novel drug candidates is a complex challenge.

Purpose of the Study:

  • To transform LLMs into specialized chemical language models for drug discovery.
  • To develop a framework for generating valid, novel, and optimized drug-like molecules.

Main Methods:

  • Supervised fine-tuning of LLMs with engineered prompts to create SmileyLlama.
  • Benchmarking SmileyLlama against existing models for molecule generation.
  • Utilizing direct preference optimization and the iMiner reinforcement learning framework.

Main Results:

  • SmileyLlama reliably generates valid and novel drug-like molecules.
  • The model predicts molecules with optimized 3D conformations and high target binding affinity.
  • SmileyLlama demonstrates user-specified property adherence.

Conclusions:

  • LLMs can be effectively adapted for specialized scientific tasks like drug discovery.
  • The SmileyLlama framework offers a powerful new tool for molecular design.
  • This approach has broad applicability in chemistry, biology, and materials science.