Table 1 The confusion matrix of EMIT with different model sizes stratified by mean mutation allele frequency

From: Development of a deep learning model for cancer diagnosis by inspecting cell-free DNA end-motifs

Groupa

Length = 64

Length = 96

Length = 128

Incorrect predictions

Correct predictions

Fisher’s exact test

Incorrect predictions

Correct predictions

Fisher’s exact test

Incorrect predictions

Correct predictions

Fisher’s exact test

EMIT-2Mb

G1

6

10

P = 0.53

4

12

P = 0.47

4

12

P = 0.29

G2

3

13

2

14

1

15

 

G3

3

12

1

14

1

14

 

G4

6

10

4

12

4

12

 

EMIT-8Mb

G1

3

13

P = 0.89

2

14

P = 1

2

14

P = 1

G2

2

14

1

15

1

15

G3

2

13

1

14

1

14

G4

4

12

1

15

1

15

EMIT-32Mb

G1

4

12

P = 0.41

2

14

P = 0.49

2

14

P = 0.74

G2

1

15

 

1

15

1

15

G3

1

14

 

0

15

0

15

G4

3

13

 

3

13

2

14

  1. aThe mean mutation allele frequency is grouped based on quantiles of 25%, 50% and 75%. G1: MAF ≤ 25%, G2: 25% < MAF ≤ 50%, G3: 50% < MAF ≤ 75%, G4: MAF > 75%.
Back to article page