Skip to Content
UW Health SMPH
American Family Children's Hospital
SHARE TEXT

Scientific Accomplishments

Probabilistically-weighted RNA-Seqalignments

 

Colin Dewey, PhD – Cancer Genetics Program

Jamie Thomson, PhD – Cell Signaling Program

  • Key to RNA-seq is accurate mapping of reads to reference genome.
  • Earlier read-mapping algorithms threw out reads that mapped to multiple sites.
  • RSEM assigns probability distribution of sites for each read and refines by expectation-maximization 
Scientific Accomplishments

 

 

Scientific Accomplishments

 

 

 

 

  • Goal is to have lowest-curve few genes with high error rate in expression measurement.

  • RSEM outperforms competitors.

 

 

 

Selected Reference (1 of 3): Li B , Ruotti V, Stewart R, Thomson J, and Dewey C. RNA-Seq gene expression estimation with read mapping uncertainty. Bioinformatics, 2010;26(4):493-500. 

 

 

Mammography - Machine Learning

 

Elizabeth Burnside, MD, PhD - Cancer Control Program

C. David Page, PhD - Cancer Genetics Program  

  • Goal is to accurately identify malignant abnormalities on mammogram.
  • Data entered in National Mammography Database standard format.
  • Ground truth malignant/benign from breast cancer registry, based on biopsy.
Burnside Scientific Accomplishments

 

 

  • Bayes net learning algorithm yields model with significantly better ROC area on held-out data than radiologists.
  • New learning algorithm SAYU performs even better.
  • Precision-recall curves are another way to plot ROC data for rare events and show room for further improvement.

Selected Reference (1 of 4): Burnside ES, Davis J, Chhatwal J, Alagoz O, Lindstrom MJ, Geller BM, Littenberg B, Shaffer KA, Kahn CE, Jr., and Page CD. Probabilistic computer model developed from clinical data in national mammography database format to classify mammographic findings. Radiology, 2009;251(3):663-72. Sponsored by NCI R01CA127379, NCI R01CA165229, NLM R01LM010921.