OpenAI’s GPT-4 language model has truly shocked scientists. It turns out that GPT-4 has remarkable potential not only in language processing but also in biological structure modeling. A study by Rutgers University revealed that GPT-4 can model biological molecules such as simple amino acids and protein structures with high precision. But interestingly, no one knows how AI gained this ability.
The OpenAI GPT-4 AI model’s high-level capability has been discovered: Modeled basic protein structures with high precision
In the study, GPT-4 was asked to create three-dimensional structure models of 20 different standard amino acids. The results were quite impressive. GPT-4 accurately predicted atomic composition, bond length and angles. However, some errors occurred when it came to ring structures and stereo-chemical configurations.
In another experiment, GPT-4 was tasked with modeling α-helix structure, a common protein structural element. Mathematical calculations were performed using the Wolfram plugin, and the model was found to be highly consistent with experimental data.
Furthermore, GPT-4 analyzed how the antiviral drug Nirmatrelvir binds with the major protease of SARS-CoV-2. The model accurately identified the amino acids involved in the binding and precisely determined the distances between atoms in the interaction.
{{user}} {{datetime}}
{{text}}