Skip to main content

Identifying genes to predict cancer radiotherapy-related fatigue with machine-learning methods

Friday, September 16, 2016 — Poster Session IV

12:00 p.m. – 1:30 p.m.
FAES Terrace


  • CA Johnson
  • K Filler
  • W Du
  • WW Lau
  • LN Saligan


Fatigue is one of the most common side effects of cancer treatments, such as radiotherapy (RT). While many factors influence the fatigue experience of patients, we hypothesize that expression of genes related to oxidative stress can be predictive of RT-related fatigue. In this study, fatigue as measured by the Functional Assessment of Cancer Therapy – Fatigue (FACT-F) and the expression level of 84 genes related to oxidative stress were assessed from 24 prostate cancer patients at baseline (pre-RT) and at RT completion. Participants were categorized into two groups, namely high fatigue (HF) and low fatigue (LF), based on a >3-point decline in FACT-F score from baseline to RT completion. In this work, we present a two-phase scheme which first selects a limited subset of genes deemed most predictive by a regularized linear regression method known as elastic net, followed by a widely used classifier, the regularized random forest, to discriminate patients having HF from LF. After selecting 16 genes out of 84 genes, the model predicted 87.5% accuracy (0.83 AUC) in cross-validation. Initial results suggest that several genes are consistently selected in the proposed scheme, such as NOS2, PRDX2 and PTGS1, showing promise as potential predictors for RT-related fatigue, and may provide information of its biologic underpinnings.

Category: Computational Biology