11/26/2003A Phylogenetic...1 A Phylogenetic Method for Identifying a Stage I Ovarian Cancer Signature in the Mass-Spectrum of Serum Proteins Jack K. Horner Science Applications International Corporation P. O. Box 3827 Santa Fe, NM
11/26/2003A Phylogenetic...2 Problem and Method Problem: ovarian cancer (OC) can currently be diagnosed only at advanced stages, when prognosis is very poor Objective: find a method that detects OC at a stage at which it can be successfully treated Method (using serum protein mass-spectra) –form a set of known cancer, and known non-cancer, spectra; call this the “reference” set –merge reference set with unknown spectrum U –apply the phylogenetic maximum parsimony method to the merged set –declare U to have the same disease state as its siblings in the tree
11/26/2003A Phylogenetic...3 Results (How well does the method work?) Correct classification rate 133 known cancer, and 95 known non-cancer, spectra ~370,000 points per spectrum Phylogenetic Analysis Using Parsimony (Ver. 4.0b10) MP heuristic search, seed from clock, 31 runs Mean.893 Std Err.003 Kurtosis Median.890 Std Dev.017 S.E. Kurt..880 Mode.890 Variance.000 Skewness.150 Minimum.860 Sum S.E. Skew..440 Maximum.930 Range.070 Valid % Confidence Interval for the mean = [.8862 to.8989]