Table 2 Top single-omic and multi-omic performance for predicting disease survival in PDAC: study validation cohorts

From: The Molecular Twin artificial-intelligence platform integrates multi-omic data to predict outcomes for pancreatic adenocarcinoma patients

Analytes

No. of training samples

No. of validation samples

No. of features

ACC (95% CI)

PPV (95% CI)

Sensitivity

Specificity

TCGA

       

Clinical and surgical pathology, DNA (SNVs, INDELs and CNVs) and RNA gene expressions,

45

109

3,024

0.96 (0.88–1.00)

0.98 (0.90–1.00)

0.95

0.98

Clinical and surgical pathology, DNA (SNVs, INDELs and CNVs), RNA gene expressions and computational pathology

45

33

3,423

0.94 (0.83–1.00)

0.95 (0.84-1.00)

0.95

0.92

DNA SNVs

72

126

351

0.94 (0.85-1.00)

0.96 (0.86-1.00)

0.95

0.94

Computational pathology

71

33

819

0.79 (0.68–0.89)

0.89 (0.78–0.99)

0.76

0.83

RNA gene expressions

57

152

1,974

0.76 (0.67–0.85)

0.80 (0.71–0.89)

0.76

0.76

DNA INDELs

56

120

43

0.72 (0.60–0.84)

0.82 (0.70–0.94)

0.68

0.77

Clinical

74

157

15

0.66 (0.57–0.75)

0.71 (0.63–0.80)

0.68

0.63

DNA CNVs

72

156

645

0.47 (0.40–0.54)

0.56 (0.49–0.63)

0.42

0.55

JHU cohort 1

       

Clinical and surgical pathology, DNA (INDELs, CNVs and SNVs), RNA gene expressions and tissue proteins

39

81

3,270

0.89 (0.83–0.95)

0.91 (0.85–0.98)

0.84

0.93

Clinical and surgical pathology, RNA gene expressions and tissue proteins

40

81

2,480

0.75 (0.69–0.82)

0.72 (0.66–0.79)

0.76

0.74

RNA gene expressions and tissue proteins

46

81

2,466

0.72 (0.66–0.79)

0.69 (0.63–0.76)

0.71

0.72

RNA gene expressions

57

81

1,963

0.68 (0.62–0.75)

0.67 (0.61–0.74)

0.63

0.72

Clinical and surgical pathology, DNA (INDELs, CNVs and SNVs) and tissue proteins

45

81

1,307

0.65 (0.59–0.72)

0.63 (0.57–0.70)

0.63

0.67

Clinical and surgical pathology, DNA (INDELs, CNVs and SNVs) and RNA gene expressions

45

81

2,767

0.60 (0.54–0.67)

0.57 (0.51–0.64)

0.63

0.58

Tissue proteins

49

81

503

0.56 (0.50–0.63)

0.53 (0.47–0.60)

0.53

0.58

DNA (INDELs, CNVs and SNVs)

56

81

790

0.51 (0.45–0.58)

0.47 (0.41–0.54)

0.45

0.56

Clinical

74

81

14

0.38 (0.32–0.45)

0.35 (0.29–0.42)

0.37

0.4

JHU cohort 2

       

Clinical and plasma proteins

41

47

255

0.98 (0.83–1.00)

0.92 (0.79–1.00)

1.00

0.97

Plasma proteins

51

47

251

0.98 (0.83–1.00)

0.92 (0.79–1.00)

1.00

0.97

Clinical, plasma proteins and plasma lipids

51

47

619

0.79 (0.63–0.94)

0.57 (0.44–0.69)

0.67

0.83

CA 19-9 presurgery

63

48

1

0.69 (0.57–0.81)

0.17 (0.04–0.40)

0.08

0.86

Plasma proteins and plasma lipids

51

47

615

0.55 (0.41–0.69)

0.30 (0.16–0.44)

0.58

0.54

Clinical

74

49

5

0.43 (0.29–0.57)

0.14 (0.02–0.26)

0.25

0.49

Clinical and plasma lipids

51

47

369

0.32 (0.20–0.44)

0.12 (0.00–0.25)

0.25

0.34

Plasma lipids

51

47

365

0.23 (0.12–0.35)

0.15 (0.03–0.27)

0.42

0.17

MGH cohort

       

Clinical and plasma proteins

51

35

259

0.91 (0.77–1.00)

0.84 (0.71–0.97)

1.00

0.84

Plasma proteins

51

35

250

0.89 (0.76–1.00)

0.80 (0.69–0.91)

1.00

0.79

Plasma proteins and plasma lipids

51

35

614

0.74 (0.61–0.87)

0.68 (0.54–0.82)

0.81

0.68

CA 19-9 presurgery

63

32

1

0.62 (0.51–0.73)

0.60 (0.52–0.68)

0.60

0.65

Clinical, plasma proteins and plasma lipids

51

35

623

0.51 (0.41–0.62)

0.47 (0.33–0.61)

0.44

0.58

Plasma lipids

51

35

365

0.49 (0.36–0.62)

0.46 (0.30–0.62)

0.69

0.32

Clinical

74

35

10

0.40 (0.29–0.51)

0.37 (0.26–0.48)

0.44

0.37

Clinical and plasma lipids

51

35

374

0.37 (0.22–0.52)

0.35 (0.20–0.49)

0.44

0.32

  1. Detailed results table for top survival models for each validation test cohorts TCGA, JHU cohort 1, JHU cohort 2 and MGH cohort. Single and multi-omic analytes predicting DS are listed in decreasing order of predictive performance for DS, arranged by ACC and PPV. For each analyte, the of number of samples available, trained and features extracted for that respective analyte are shown. The predictive performance for each analyte is based on the best-performing model. Analytes within each validation set are listed in decreasing order of survival accuracy. Bold ACC and PPV values indicate the best-performing analyte within each cohort.
Back to article page