Evaluation of GenAI in Biomedical Knowledge Discovery

Abstract

This paper evaluates the power of GenAI and its generated results and compares them with the more traditional way of computational reasoning for biomedical knowledge discoveries. The experiments conducted in this research focus on the problem of discovering activated/inhibited enzymes involved in diseases, but the same idea of evaluating GenAI can be applied to any problem domain. The traditional way of knowledge discovery through the power of computational logic reasoning is now challenged by the rather unexpected power of LLM and transformer architecture, such as ChatGPT, in almost any discipline of science and humanities. The results from this research are interesting: ChatGPT disappoints in biomedical knowledge discoveries because of the lack of precision and absence of reasoning. However, it is feasible to define pathways, when using GenAI, which can remedy current ChatGPT shortcomings.