How autonomous AI agents will change cancer research
More time for the essential
In a publication in the journal “Nature Cancer”, researchers from Else Kröner Fresenius Center (EKFZ) for Digital Health at TU Dresden and Genentech, a member of the Roche group, explain how autonomous AI models will change workflows in cancer research and beyond in the future.
Artificial intelligence (AI) and deep learning are already supporting researchers in numerous areas. Until now, these models have only been able to solve specific tasks, for which they require precise specifications and guidance from scientists. Biomedical research, such as the development of new cancer therapies, usually involves complex multi-step workflows. These include researching, planning and conducting experiments, followed by evaluation and interpretation of data. So far, AI has only been able to help with individual steps such as data analysis or modeling. With the introduction of large language models (LLMs) such as ChatGPT, which work on the basis of natural human language, additional scientific tasks such as literature research, hypothesis generation and planning experiments can now be supported by AI. These new models not only understand pure text, but also images, videos and structured data such as tables and flowcharts. The concept of AI agents, language models such as ChatGPT, could in future access any software itself and use it to solve a task – something that could previously only be performed by scientists. The models are also becoming increasingly better at learning independently, reflecting on their knowledge and solving new problems. In their publication in the journal “Nature Cancer”, the researchers describe how these advancements will change scientific work in cancer research in the near future.
AI agents can support researchers
Autonomous AI models based on LLMs that learn and reflect independently could work seamlessly with researchers in the future. This would speed up the entire development process in cancer research, from literature research and project planning to the modeling of potential drugs and the design of clinical trials.
Simplify complex workflows
The models simplify time-consuming biomedical workflows by automating multi-step tasks and enabling efficient collaboration between specialized AI systems. Identifying new targets for cancer drugs involves extensive literature research. Complex modeling of the 3D structure of a protein or drug has often been the subject of an entire PhD thesis. New AI agents with Internet access, on the other hand, can read hundreds of publications and examine numerous different 3D structures within a few minutes.
More time for creative ideas and strategic decisions
Even if the systems could work increasingly autonomously, human researchers will still continue to monitor them. They will guide the systems and check individual steps and results. The aim is to automate detailed, time-consuming routine work. This will leave scientists more time for creative new ideas and strategic decisions.
“These new systems will significantly change and accelerate biomedical research. At the same time, scientists must also be aware of the ethical and safety-related consequences. It is our task to use AI responsibly and to define the necessary framework conditions. Then, these AI systems will be a valuable addition and support for advancing research, better understanding diseases and finding suitable therapeutic approaches,” says Prof. Jakob N. Kather, Professor of Clinical Artificial Intelligence at TUD Dresden University of Technology.
Original publication
Other news from the department science
Most read news
More news from our other portals
See the theme worlds for related content
Topic world Digitalization in the laboratory
The topic world Digitalization in the lab presents innovations and trends from digital data systems (ELN, LIMS) to laboratory robots and networked devices (IoT) to AI and machine learning.
Topic world Digitalization in the laboratory
The topic world Digitalization in the lab presents innovations and trends from digital data systems (ELN, LIMS) to laboratory robots and networked devices (IoT) to AI and machine learning.
