Format

Send to

Choose Destination
Pain. 2019 May 15. doi: 10.1097/j.pain.0000000000001616. [Epub ahead of print]

Machine-learned analysis of the association of next-generation sequencing based genotypes with persistent pain after breast cancer surgery.

Author information

1
Institute of Clinical Pharmacology, Goethe-University, Theodor-Stern-Kai 7, 60590 Frankfurt am Main, Germany.
2
Faculty of Biological Sciences (FB15), Goethe-University, Max-von-Laue-Strasse 9, 60438 Frankfurt am Main, Germany.
3
Institute for Molecular Medicine Finland (FIMM), HiLIFE, University of Helsinki, Helsinki, Finland.
4
Division of Pain Medicine, Dept. of Anaesthesiology, Intensive Care and Pain Medicine, University of Helsinki and Helsinki University Hospital, Helsinki, Finland.
5
Fraunhofer Institute of Molecular Biology and Applied Ecology - Project Group Translational Medicine and Pharmacology (IME-TMP), Theodor-Stern-Kai 7, 60590 Frankfurt am Main, Germany.

Abstract

Cancer and its surgical treatment are among the most important triggering events for persistent pain, but additional factors need to be present for the clinical manifestation, such as variants in pain-relevant genes. In a cohort of 140 women undergoing breast cancer surgery, assigned based on a three-year follow-up to either a persistent or non-persistent pain phenotype, next generation sequencing was performed for 77 genes selected for known functional involvement in persistent pain. Applying machine learning and item categorization techniques, 21 variants in 13 different genes were found to be relevant to the assignment of a patient to either the persistent pain or the non-persistent pain phenotype group. In descending order of importance for correct group assignment, the relevant genes comprised DRD1, FAAH, GCH1, GPR132, OPRM1, DRD3, RELN, GABRA5, NF1, COMT, TRPA1, ABHD6, and DRD4, of which one in the DRD4 gene was a novel discovery. Particularly relevant variants were found in the DRD1 and GPR132 genes, or in a cis-eCTL position of the OPRM1 gene. Supervised machine learning based classifiers, trained with 2/3 of the data, identified the correct pain phenotype group in the remaining 1/3 of the patients at accuracies and areas under the receiver operator characteristic curves of 65 - 72 %. When using conservative classical statistical approaches, none of the variants passed α-corrected testing. The present data analysis approach, using machine learning and training artificial intelligences, provided biologically plausible results and outperformed classical approaches to genotype phenotype association.

Supplemental Content

Full text links

Icon for Wolters Kluwer
Loading ...
Support Center