ChatGPT in Drug Discovery and Development: How do you use it?

ChatGPT is a well-known large language model (LLM) and is often used in our daily life and business scenes. Its application is expanded from a simple text to a picture and its potential is hard to imagine now. ChatGPT or other LLM could help acceleration of drug discovery and development as well. The issue would be how we can use the technology. It can assist us on target identification, drug design and so on, depending on the way we take advantage of LLM.

In pharmaceutical industry, ChatGPT and other LLM is used in a customized manner.¹⁾ ChatGPT itself is also applied to develop a novel strategy for drug discovery as well.²⁾

Taking the ACS paper for example, the author demonstrated a case study with anticocaine addiction drug development by the use of ChatGPT. They thoroughly took advantage of ChatGPT to merge it to usual drug discovery process.

The authors give it a key role as the facilitator of steering researchers for the creation of a drug candidate. Synergic discussion between human and AI enabled bigdata-assisted rational and productive drug development.

The impact and future possibilities of LLM in drug target discovery, drug discovery and development are recently summarized in a short review.⁴⁾ However, it is highly risky if you depend on ChatGPT. There are still limitations on confidently apply ChatGPT on drug discovery and development. You would understand the issues looking through this paper³⁾ and using current ChatGPT with suspicions.

First, ChatGPT is not always transparent and explainable. The reference of the outcome is often unavailable or on an unreliable source and hence it’s hard to utilize it as evidence to make a scientifically rational progress.

The second problem comes from the dataset of ChatGPT. ChatGPT is not trained for pharmaceutical field. It’s optimized for the needs of sufficiently wide range. It sometimes complicates the generated text because the wordings are not always suitable for science but for our daily conversation.

The last issue is the training and optimization strategy. Development of LLM enable us to use an optimized model for drug discovery. In order for accuracy, we need to figure out a requisite dataset for training the AI and apply it for a specific project. It is because each drug discovery process faces with specific issue to solve, which includes the problems that no one has ever encountered.

In principle, it is possible to produce a general LLM for drug discovery but highly sophisticated model is hard to establish. This is one of the crucial reasons why drug discovery companies develop customized ChatGPT.

Still, LLGs including ChatGPT is useful if you make clever use of its advantage, knowing the necessity of compensating for the weakness and risks. It is smart to use it as a tool to rapidly gather information related to the matter of interest. It is also useful as a facilitator or observer from the objective point of view. The technology is significant for the aid of research on drug discovery and development. We need to run a trial and error to figure out our own application to maximize its benefit.

1) https://doi.org/10.1038/s41587-023-01788-7
2) https://doi.org/10.1021/acs.jcim.3c01429
3) https://doi.org/10.1016/j.omtn.2023.08.009
4) https://doi.org/10.3389%2Ffphar.2023.1194216