Table 2.

Samples misclassified by the 5-gene model using FLDA and/or SVM


Sample

FLDA

SVM

PLS3
C010.1.Th1*  0.02 ± 0.29  0.12 ± 0.04  0.007  
C021.2.Th1  -1.40 ± 0.40   -0.03 ± 0.07   0.009  
C019.4.UT  -0.49 ± 0.27   -0.02 ± 0.05   0.007  
C052.1.UT   0.12 ± 0.16  -0.10 ± 0.03   0.007  
C078.1.UT   0.50 ± 0.15  -0.15 ± 0.03   0.006  
C045.1.UT   0.05 ± 0.15  -0.22 ± 0.03   0.006  
S108.1.05   2.73 ± 0.56   -0.04 ± 0.09  0.033  
S161.1.10*  -6.35 ± 0.40  -0.56 ± 0.05  0.001  
S157.3.15   0.74 ± 0.14   -0.12 ± 0.03  0.004  
S160.1.15*  -2.19 ± 0.33  -0.25 ± 0.05  0.075§ 
S137.1.27*  -1.77 ± 0.25  -0.16 ± 0.03  0.001  
S142.1.30*  -1.55 ± 0.36  -0.08 ± 0.08  0.700§ 
S156.1.30*  -0.61 ± 0.21  -0.12 ± 0.04  0.001  
S140.1.39   0.75 ± 0.17   -0.04 ± 0.04  0.004  
S133.1.46*  -1.30 ± 0.21  -0.15 ± 0.03  0.002  
S130.1.48   -1.78 ± 0.30  0.10 ± 0.07   0.939§ 
S159.1.70   0.28 ± 0.10   -0.15 ± 0.03  0.035  
S132.1.84*  -2.31 ± 0.25  -0.24 ± 0.03  0.001  
S218.1.87*
 
-0.66 ± 0.20
 
-0.44 ± 0.07
 
0.218§
 

Sample

FLDA

SVM

PLS3
C010.1.Th1*  0.02 ± 0.29  0.12 ± 0.04  0.007  
C021.2.Th1  -1.40 ± 0.40   -0.03 ± 0.07   0.009  
C019.4.UT  -0.49 ± 0.27   -0.02 ± 0.05   0.007  
C052.1.UT   0.12 ± 0.16  -0.10 ± 0.03   0.007  
C078.1.UT   0.50 ± 0.15  -0.15 ± 0.03   0.006  
C045.1.UT   0.05 ± 0.15  -0.22 ± 0.03   0.006  
S108.1.05   2.73 ± 0.56   -0.04 ± 0.09  0.033  
S161.1.10*  -6.35 ± 0.40  -0.56 ± 0.05  0.001  
S157.3.15   0.74 ± 0.14   -0.12 ± 0.03  0.004  
S160.1.15*  -2.19 ± 0.33  -0.25 ± 0.05  0.075§ 
S137.1.27*  -1.77 ± 0.25  -0.16 ± 0.03  0.001  
S142.1.30*  -1.55 ± 0.36  -0.08 ± 0.08  0.700§ 
S156.1.30*  -0.61 ± 0.21  -0.12 ± 0.04  0.001  
S140.1.39   0.75 ± 0.17   -0.04 ± 0.04  0.004  
S133.1.46*  -1.30 ± 0.21  -0.15 ± 0.03  0.002  
S130.1.48   -1.78 ± 0.30  0.10 ± 0.07   0.939§ 
S159.1.70   0.28 ± 0.10   -0.15 ± 0.03  0.035  
S132.1.84*  -2.31 ± 0.25  -0.24 ± 0.03  0.001  
S218.1.87*
 
-0.66 ± 0.20
 
-0.44 ± 0.07
 
0.218§
 

Patient samples are sorted by their tumor burden. Columns LDA and SVM contain average scores obtained by these samples ± standard deviation over the approximately 100 times that these samples were tested during 1000 random resamplings. The column labeled PLS3 displays the expression level of PLS3 (in terms of its ratio to the housekeeping gene relative to that in the Stratagene reference RNA sample).

*

Samples that were misclassified by both methods.

Scores assigned these samples to incorrect class (positive average scores for controls and negative average scores for patients).

Two control samples, C021.2.Th1 and C019.4.UT, were marginally classified by SVM, even though the average score turned out to be negative, and were considered to be misclassified.

§

Misclassified patients with high levels of PLS3 (with z score more than 5, relative to controls). These high-PLS3 patients may deserve special treatment, because high levels of PLS3 alone are very unusual for healthy individuals.

Close Modal

or Create an Account

Close Modal
Close Modal