1 / 1

polyomx

www.polyomx.org. Using Single Nucleotide Polymorphisms to Predict Radiation Toxicity in Prostate Cancer Patients Treated with Conformal Radiotherapy. Wang Y 1,2 , Damaraju S 1,4,5 , Cass CE 1,4,5 , Murray D 4,5 , Fallone G 4,5 , Parliament M 4,5 and Greiner R 1,2,3

laura-kane
Télécharger la présentation

polyomx

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript


  1. www.polyomx.org Using Single Nucleotide Polymorphisms to Predict Radiation Toxicity in Prostate Cancer Patients Treated with Conformal Radiotherapy Wang Y1,2, Damaraju S1,4,5, Cass CE1,4,5, Murray D4,5, Fallone G4,5, Parliament M4,5 and Greiner R1,2,3 PolyomX Program1, Department of Computing Science2 and Alberta Ingenuity Centre for Machine Learning3 and Oncology4, U of A and Cross Cancer Institute5 GOAL: Predict toxicity in patients' treated with radiation • Used Machine Learning (ML) to build classifiers … • over 51 SNPs in genes encoding DNA damage, recognition/repair/response and clinical radiation toxicity • in retrospective cohort of patients (n=82) treated for prostate cancer with conformal radiotherapy (3DCRT) Definitions: • SNP (Single Nucleotide Polymorphism): • commonly occurring genetic variations • may affect an individual's susceptibility to disease or response to particular treatment by altering the expression of the gene in which it occurs. - - • Method • For each patient • SNPs = features (independent variables) • response to treatment = class label (dependent variable) • RTOG>2 after 90 days == “negative” • Learn classifiers: • "J48" decision tree • "KStar" nearest-neighbor • Filtering (for each classifier) • Used information gain to rank the SNPs • Considered classifiers based on top k SNPs, for k=1, 2, … • To identify the best classification system: • Use 10-fold cross-validation to estimate predictive accuracy of each classifier with each feature subset • Use permutation test (4000 trials) to test significance of our results • Information Content • information about class variable provided by attribute variable • Typically better to use attribute with high information content • Results • 70-80% prediction accuracy, using SNPs (rank order): • XRCC3 (A>G, 5’ UTR Nt 4541), • CYP2D6*4 (G>A, Splicing defect), • BRCA2 (A>G, K 1132 K), • MLH1 (C>T, V 219 I), • BRCA1 (A>G, R 356 Q), • RAD51 (G>T, 5’ UTR Nt 172), • BRCA2 (A>G, S 455 S), • BRCA2 (C>A, N 289 H), • BRCA2 (A>G, D 991 N). • 4,000-trial permutation test: • significance at p<0.05 level, • for both J48 and KStar - - - • 0 if attribute “a” is NOT correlated with class “c” • Positive if correlated • K-fold Cross-Validation: used to estimate quality of model • Example (K=3) • Radiation toxicity: • Patient treated with conformal radiotherapy (3DCRT) • Every visit: MD gives patient RTOG score ( 0 - 5 ) • Toxicity if • RTOG > 2 (constant bleeding ) after 90 days • Machine Learning • Classifier = program that returns “label” for novel instance • Learner = computer program that produces classifier, based on patterns found in training data Decision Tree: • Conclusion • Machine Learning techniques can be used for SNP data analyses and clinical treatment outcome prediction • Machine Learning helps …in discriminating between • populations according to SNP data towards identifying predictive SNPs for use in radio-genomics in the near • future. • Permutation test: • Randomly rearrange LABELS of data (so should be NO signal!) • Run same learning algorithm, observe classification accuracy • . KStar: nearest neighbor using a generalized distance function based on transformations - -

More Related