Random_Intensity_60_4a: : 60 Cases, 60 Controls, 300 Peaks

Brian T. Luke (lukeb@ncifcrf.gov)
Return to Description of Tables
PDF Version

BMDK Analysis

22 peaks selected as putative biomarkers by the 10 methods within BMDK

Peak catboot student dtgini dtinfg nnfeat chisq kruswal kolsmir extreme vartest
27   3     1   3 3   3
28                 3  
36     4     1   3    
43 5 2     2   1 3   2
82 1                  
83       3   5     1  
99     1 1       1    
117         3     3    
126           3        
127 4                  
139           5     1  
146     5 5            
166                 3  
209                 3  
220     2 2            
234           2        
245                 3  
247   1 3 4 5 3 2 1   1
257   5         5     5
262 3                  
269   4     4   4     4
286 2                  

Peaks used in each of the best distance-dependent 6-nearest neighbor classifiers

Peak 1-ad 2-ad 3-ad 1-rd 2-rd 3-rd 2-cr 3-cr 2-sd 3-sd
27                    
28               X    
36                 X X
43           X        
82               X    
83     X           X X
99     X     X       X
117 X X   X X   X      
126           X        
127                    
139                    
146     X         X    
166                    
209                    
220                    
234                    
245                    
247                    
257                    
262         X   X      
269   X                
286                    

Sensitivity, specificity, %undetermined, and quality (sensitivity + specificity - %undetermined) for each of the best distance-dependent 6-nearest neighbor classifiers using any of the 22 putative biomarkers.

Metric 1-ad 2-ad 3-ad 1-rd 2-rd 3-rd 2-cr 3-cr 2-sd 3-sd
Sens 68.3 77.6 77.1 69.6 68.6 77.5 66.1 73.1 68.6 70.6
Spec 61.7 63.3 75.0 60.3 76.9 76.2 71.9 71.2 73.7 78.7
%Undet 0.0 1.7 16.7 5.0 14.2 31.7 5.8 13.3 10.0 18.3
Quality 130.0 139.3 135.4 125.0 131.4 122.0 132.2 130.9 132.3 131.0

Sensitivity, specificity, %undetermined, and quality (sensitivity + specificity - %undetermined) for each of the best distance-dependent 6-nearest neighbor classifiers using any of the 22 putative biomarkers with the caveat that %Undetermined cannot exceed 5.0%.

Metric 1-ad 2-ad 3-ad 1-rd 2-rd 3-rd 2-cr 3-cr 2-sd 3-sd
Sens 68.3 77.6 71.2 69.6 None None 75.0 None 75.4 None
Spec 61.7 63.3 63.3 60.3 None None 53.7 None 57.9 None
%Undet 0.0 1.7 0.8 5.0 None None 5.0 None 5.0 None
Quality 130.0 139.3 133.7 125.0 None None 123.7 None 128.3 None

Fingerprint Analysis

Sensitivity, specificity and quality (sensitivity + specificity) for the best and 200th best decision tree constructed from any of the 300 peak intensities.  The evolutionary programming search used a population size of 200 and ran for 400 generations.  A decision node became a terminal node when it contained 1% (no samples) or 4% (2 samples) of a given State.

Metric 1% 1% 4% 4%
1st 200th 1st 200th 1st 200th 1st 200th
Sensitivity 93.3 90.0 88.3 83.3 91.7 85.0 83.3 83.3
Specificity 83.3 83.3 88.3 88.3 85.0 88.3 88.3 85.0
Quality 176.7 173.3 176.7 171.7 176.7 173.3 171.7 168.3

Sensitivity, specificity and quality (sensitivity + specificity) for the best and 200th best medoid classifier algorithm in each of the two runs using 5-, 6-, and 7-peak intensities from the set of 300.  The evolutionary programming search used a population size of 400 and ran for 800 generations with the requirement that there are at most 40 Case-cells and 40 Control-cells.

Metric 5-Features 5-Features 6-Features 6-Features 7-Features 7-Features
1st 200th 1st 200th 1st 200th 1st 200th 1st 200th 1st 200th
Sens 100.0 100.0 90.0 83.3 100.0 100.0 93.3 85.0 100.0 100.0 91.7 86.7
Spec 93.3 83.3 100.0 100.0 93.3 85.0 100.0 100.0 95.0 90.0 100.0 100.0
Quality 193.3 183.3 190.0 183.3 193.3 185.0 193.3 185.0 195.0 190.0 191.7 186.7

(Last updated 4/21/07)