Skip to main content

Table 2 Performance of DNN models

From: A deep neural network based regression model for triglyceride concentrations prediction using epigenome-wide DNA methylation profiles

Data

Evaluation Metric

Cutoffsa

Min

1st quartile

Mean

Median

3rd quartile

10kCpGs

1kCpGs

1

RMSE

88.5 (26.3)

88.8 (25.6)

89.3 (25.7)

89.0 (27.3)

88.8 (26.1)

89.2 (25.9)

89.8 (26.4)

Cor

0.19 (0.05)

0.27 (0.08)

0.19 (0.09)

0.14 (0.11)

0.11 (0.10)

0.24 (0.02)

0.14 (0.11)

2

RMSE

48.5 (14.4)

48.4 (14.7)

47.4 (13.7)

48.5 (14.3)

47.5 (13.8)

48.6 (12.9)

48.8 (13.0)

Cor

0.23 (0.13)

0.10 (0.29)

0.29 (0.07)

0.14 (0.19)

0.29 (0.07)

0.20 (0.11)

0.10 (0.14)

3

RMSE

48.5 (4.7)

48.7 (4.8)

48.5 (4.5)

48.1 (3.5)

48.6 (4.6)

48.2 (5.0)

48.5 (5.3)

Cor

0.17 (0.07)

0.18 (0.08)

0.22 (0.13)

0.20 (0.12)

0.19 (0.08)

0.17 (0.06)

0.16 (0.04)

  1. aThe selected CpG sites with interindividual variability greater than or equal to different cutoffs of DNAm values (minimum [no filtering], first quartile, second quartile, mean, and third quartile) as well as the top 10,000 CpG sites (10kCpGs) and top 1000 CpG sites (1kCpGs)