By applying ML to NCI’s The Cancer Genome Atlas (TCGA), researchers are characterizing molecular “fingerprints” for major cancer types. What if you could tap into these molecular profiles (e.g., gene expression, DNA methylation, microRNA, copy number variation, and mutations) and apply them to your own omics data? Thanks to a new study, you have some important new tools for doing just that.

An international study, funded by NCI, gives you an online resource with 737 publicly available top-performing predictive models for 26 cancer cohorts, data types, and training algorithms.

This resource can help you classify your non-TCGA patient samples based on findings from TCGA’s molecular cancer subtypes. You can look for molecular variations associated with a variety of diverse cancers.

Senior author, Dr. Peter W. Laird, from Van Andel Institute in Michigan, said, “With this study, we’re beginning to bridge an important gap between cancer research and clinical practice. Using what we know about molecular subtypes in existing cancer cohorts, we can apply ML and TCGA’s subtype labels to newly diagnosed patients in the clinic. This is giving us a strong foundation for developing new clinical tests—to help in future genomic studies and in targeting participants to the most appropriate clinical trial.

NCI’s Division of Cancer Biology funded this work.