A team of scientists from the Norwich Medical School at the University of East Anglia (UEA) is calling into question some of the findings relating to machine learning in the highly cited 2020 Nature paper, ‘Microbiome analyses of blood and tissues suggest cancer diagnostic approach’.
Led by Dr Abraham Gihawi and Prof Daniel Brewer of the Norwich Medical School, the team reanalysed work by another group claiming that microbial DNA found in blood and tissue has a near-perfect ability to identify where in the body a cancer is located.
The UEA researchers found that many of the bacteria reported in the original study were identified incorrectly and that errors in machine learning practices lead to tests performing much better than they would have done otherwise.
This impacts many of the studies and findings that have cited this work and raises issues regarding the use of machine learning and artificial intelligence more generally.
There needs to be a more thorough investigation into the role that microbes play in cancer in ongoing research.
Published: