Table 2 Performance of polygenic prognostic scores derived from different approaches in the TCGA cohort

From: Integration of pathologic characteristics, genetic risk and lifestyle exposure for colorectal cancer survival assessment

Method

Parametera

Weightb

NSNPc

All follow-up

3-year follow-up

5-year follow-up

HR (95% CI)d

Pd

HR (95% CI)d

Pd

AUCe

HR (95% CI)d

Pd

AUCe

C + T

0.001

Meta

300

1.97 (1.56, 2.50)

1.70E−08

2.07 (1.56, 2.75)

4.92E−07

0.653

2.04 (1.59, 2.61)

1.82E−08

0.635

C + T

1.00E−04

Meta

34

1.47 (1.17, 1.85)

0.001

1.45 (1.11, 1.91)

0.007

0.606

1.45 (1.14, 1.84)

0.002

0.541

C + T

1.00E−05

Meta

5

1.18 (0.96, 1.45)

0.116

1.19 (0.94, 1.51)

0.156

0.575

1.18 (0.95, 1.47)

0.123

0.531

LASSO

0.01

Meta

287

1.95 (1.54, 2.46)

2.41E−08

2.08 (1.57, 2.76)

4.10E−07

0.635

2.04 (1.59, 2.61)

1.63E−08

0.633

LASSO

0.01

Penalty

287

1.97 (1.55, 2.50)

3.08E−08

2.10 (1.58, 2.77)

2.15E−07

0.670

2.07 (1.61, 2.67)

1.88E−08

0.643

RSFf

Optimal

Meta

287

1.99 (1.57, 2.52)

1.76E−08

2.06 (1.55, 2.75)

8.22E−07

0.659

2.05 (1.59, 2.63)

2.10E−08

0.652

CoxBoost

Optimal

Meta

265

1.97 (1.55, 2.49)

2.27E−08

2.09 (1.58, 2.77)

2.27E−07

0.657

2.07 (1.61, 2.66)

1.02E−08

0.641

CoxBoost

Optimal

Boosting

265

1.64 (1.31, 2.06)

1.75E−05

1.74 (1.33, 2.28)

5.92E−05

0.665

1.67 (1.31, 2.12)

2.80E−05

0.555

  1. PPS polygenic prognostic score, TCGA The Cancer Genome Atlas, SNP single nucleotide polymorphism, C + T clumping and P value thresholding, LASSO least absolute shrinkage and selection operator, RSF random survival forest, HR hazard ratio, 95% CI 95% confidence interval, SD standard deviation, ROC receiver operating characteristics.
  2. aParameters for SNPs section: P value for C + T method; lambda for LASSO method; optimal AUC for RSF method; optimal boosting steps for CoxBoost method.
  3. bWeight for PPS construction, derived from meta-analysis or penalized/boosted regression.
  4. cNumber of SNPs in the derivation stage.
  5. dHR (95% CI) per SD, derived from cox regression model with the adjustment of sex, age, stage and top 10 principal components. The P value is two-sided.
  6. eArea under the time-dependent ROC curve.
  7. fThe optimal PPS was highlighted in bold.
Back to article page