The use of metabolomics in INCISIVE: a preliminary work

A new study has been published by the Laboratoy of Computing, Medical Informatics and Biomedical Imaging Technologies of the Aristotle University of Thessaloniki (AUTH), entitled “Combining machine learning and network analysis pipelines: the case of microbiome and metabolomics data in colorectal cancer”.

A new study of the Aristotle University of Thessaloniki has analysed samples of intestinal microbiome and metabolites, from healthy individuals (HE) and patients with adenomas (AD or colorectal carcinomas (CRC), to examine the correlations of specific metabolites and microbes with the occurrence of cancer with the NetAn methods.

The study has shown that the distinction of HE and AD is successful, while CRC distinction presented lower success. The fewer samples of the carcinoma class had a strong negative effect on the performance of the models. For a set of samples where the number of samples was equal to the other classes, the features selected with NetAn may lead to better predictions than a classic feature selection method.

The comparisons indicated that by a large majority, the models produced better results in the analysis of the subsets compared to the complete set of features. The distinction of HE and AD is successful (average values MI: HE~77%, AD~78% and ME: HE~80%, AD~81%), but the distinction of CRC presented lower success regardless of the model used, the dataset or the subset applied. In both the MI and ME datasets, the prediction using features from NetAn showed better results in distinguishing HE from AD or CRC.

This study is a preliminary work for INCISIVE regarding the use of metabolomics that will help the projects’ research of delivering AI tools and a large data repository, as well as many other challenges related, to the explorations in cancer imaging.

You can find the paper in our publications.

Want to know more about the project? Keep following INCISIVE on Linkedin and Twitter to keep updated with its next developments.