Using automated text analysis and Eve, the robot scientist, the University of Cambridge-led researchers recently analyzed over 12,000 research studies on breast cancer biology.
A New Scientist report specified that the researchers used the said combination to "semi-automate" the reproduction process of research results. The problem of the absence of "reproducibility" is one of the biggest crises that face modern science.
After slimming down the set to 74 papers of high scientific interest, less than a third or 22 papers were discovered to be reproducible. In two cases, Eve was found to have made "serendipitous discoveries."
A successful experiment is where another scientist from a different laboratory under the same conditions can attain the same result. However, over 70 percent of scientists have attempted and failed to reproduce experiments of another scientist, and more than 50 percent have been unable to reproduce some of their experiments. Such a failure is called the "reproducibility crisis."
Use of Computer-Robotic System to Perform Scientific Experiments
The study published in the Royal Society Interface journal showed that it is plausible to use artificial intelligence or robotics to help solve the reproducibility crisis.
According to Professor Ross King from Cambridge's Department of Chemical Engineering and Biotechnology, who led the study, good science depends on being reproducible. Otherwise, he elaborated, "the results are meaningless."
This is especially crucial in biomedicine, King said, adding, "If I am a patient," and he read about a promising new potential treatment, but the results are not reproducible, he wouldn't know what to believe. People would possibly lose trust in science, the professor continued. Several years back, King invented Eve, a computer-robotic system that employs techniques from AI to perform scientific experiments.
The developer said that one of the big advantages of employing machines to do science is that they are more accurate and record details more precisely than humans can do. This makes them well-suited to the task of trying to reproduce scientific results.
Combining Robotics and AI to Deal with Reproducibility Crisis
As part of a DARPA-funded project, King, together with his colleagues from the United Kingdom, United States, and Sweden, designed an experiment that utilizes a combination of robotics and AI to help solve the reproducibility crisis by getting computers to read scientific papers, then understand them, and getting Eve to try to reproduce the experiments.
For this current study, a similar ScienceDaily report specified that the cancer literature is enormous, although no one ever is doing the same thing twice, which makes reproducibility a major issue.
'Robot scientist' Eve finds that less than one third of scientific results are reproducible https://t.co/LPSFMLO0BN #AI #Sciences
— Nanowerk (@Nanowerk) April 6, 2022
Given the huge amount of money spent on cancer research and the sheer number of individuals affected by cancer globally, it is an area with an urgent need to enhance reproducibility, explained King.
From an initial set of over 12,000 published scientific research, the scientists used automated text mining methods to extract statements associated with a change in gene expression in response to drug treatment in breast cancer. Then, from this set, there was a selection of 74 papers.
According to King, this work shows that automated and semi-automated approaches could be an essential mechanism to help deal with the reproducibility crisis and that reproducibility needs to be a standard part of the scientific procedure.
Related information about breast cancer is shown on ABC News Australia's YouTube video below:
Check out more news and information on Artificial Intelligence and Breast Cancer in Science Times.
