How artificial intelligence is transforming drug discovery—and why real-world testing still matters

What is it about?

Artificial intelligence (AI), and especially deep learning, is rapidly changing how new drugs are discovered. These methods can analyze large datasets, identify promising drug candidates, and predict how molecules will behave—much faster than traditional approaches. In this study, we conducted a systematic review of how deep learning is used across the drug discovery process, from early-stage screening to predicting biological activity and safety. We examined a wide range of models, including neural networks and other advanced AI techniques. A key focus of the review is in vivo validation—testing predictions in real biological systems (such as animal models). While many AI models show strong performance in computational or laboratory (in vitro) settings, fewer studies validate these results in living organisms. The review highlights both the strengths of deep learning in accelerating drug discovery and the current gaps in translating predictions into real-world biological outcomes.

Featured Image

Why is it important?

Developing new drugs is time-consuming, expensive, and has a high failure rate. AI offers the potential to dramatically improve this process by identifying better candidates earlier and reducing unnecessary experiments. This work is important because it provides a critical evaluation of how deep learning is actually being applied in drug discovery, rather than just its theoretical potential. By emphasizing the lack of in vivo validation in many studies, the review highlights a key challenge: ensuring that AI predictions are reliable in real biological systems. This is essential for translating computational advances into clinically useful therapies. The study also helps guide future research by identifying best practices and areas where improvements are needed, supporting more robust and trustworthy AI-driven drug development.

Perspectives