Artificial intelligence utility for drug development: ChatGPT and beyond

Gurwitz, David; Shomron, Noam

doi:10.1002/ddr.22121

Cited by 2 publications

(1 citation statement)

References 27 publications

(23 reference statements)

Supporting

Mentioning

Contrasting

Order By: Relevance

“…More recently, generative AI language models, particularly Chat Generative Pre-trained Transformer (ChatGPT) [16][17][18], have garnered substantial interest with applications including performance of autonomous research [19], aiding drug discovery [20][21][22], bioinformatic analysis [23,24], among others [25][26][27]. Unlike AlphaFold2 and RoseTTAFold, ChatGPT is trained on natural language datasets and its neural network architecture is tailored towards understanding and generating natural language text rather than structural modelling.…”

Section: Introductionmentioning

confidence: 99%

Generative artificial intelligence performs rudimentary structural biology modeling

Ille,

Markosian,

Burley

et al. 2024

Preprint

View full text Add to dashboard Cite

Generative artificial intelligence (AI) is being increasingly used by researchers in numerous fields of study, including the biological sciences. To date, the most commonly used tool grounded in this technology has been Chat Generative Pre-trained Transformer (ChatGPT). While ChatGPT is typically applied for natural language text generation, other application modes include coding and mathematical problem-solving. We have recently reported the ability of ChatGPT to interpret the central dogma of molecular biology and the genetic code. Here we explored how ChatGPT performs rudimentary structural biology modelling in order to gain evaluative insight. We prompted ChatGPT to model 3D structures for the 20 standard amino acids as well as an α-helical polypeptide chain, with the latter involving incorporation of the Wolfram plugin for advanced mathematical computation. For amino acid modelling, distances and angles between atoms of the generated structures in most cases approximated to around experimentally-determined values. For α-helix modelling, the generated structures were comparable to that of an experimentally-determined α-helical structure. However, both amino acid and α-helix modelling were sporadically error-prone and increased molecular complexity was not well tolerated. Despite current limitations, our findings show the capability of generative AI to perform basic structural biology modelling with atomic-scale accuracy. These results provide a precedent for the potential use of generative AI in structural biology as this technology continues to advance.

show abstract