Generative Artificial Intelligence (AI) focuses on creating new data rather than just analyzing existing data or making predictions based on it. In other words, generative AI can generate new content such as images, music, text, or even video based on patterns it has learned from existing data.
Generative AI is a type of machine learning that enables machines to improve their performance without explicit programming. This technique typically uses neural networks, algorithms designed to recognize patterns in data. Generative AI has many potential applications, such as generating realistic images and videos for entertainment, creating new music, improving medical diagnoses by generating more accurate medical images, and assisting in drug discovery by generating new molecules with specific properties.
According to Nature, Generative AI has the potential to accelerate the development of new drugs by generating novel molecules with specific properties. Antibodies are crucial weapons in the immune system’s fight against infection. They are widely used in biotech to develop drugs for various conditions, including breast cancer and rheumatoid arthritis. However, developing and improving antibodies can be a laborious process that involves a lot of screening, which is where generative AI tools can help.
The research team, led by biochemist Peter Kim and computational biologist Brian Hie at Stanford University in California, used a protein language model similar to the significant language models used by ChatGPT but is trained on tens of millions of protein sequences instead of vast volumes of text.
Despite being trained on just a few thousand antibody sequences out of the nearly 100 million protein sequences it had learned from, the model was able to suggest a small number of mutations for antibodies that boosted their ability to recognize and block the proteins used by viruses like SARS-CoV-2 and ebolavirus to infect cells.
Some of the suggested changes occurred outside the regions of the antibody that interact with its target, which is typically the focus of engineering efforts. This suggests that the model can reach for information that is not obvious to antibody engineering experts and could help researchers design new drugs for molecular targets that have resisted conventional antibody-design approaches.
While the study represents a significant step forward in antibody design, some researchers believe that generative AI can also create new antibodies that recognize a specific target. This could be particularly useful in developing drugs for conditions such as neurologic disorders, heart disease, and certain types of cancer.
However, one of the challenges in designing completely new antibodies is that their ability to recognize a particular target depends on floppy loops in the antibody structure, which can be challenging to model with AI. Nevertheless, recent studies have shown promising results in this area, and researchers are optimistic that generative AI will play an increasingly important role in developing new antibody drugs in the future.
