Table 2 Top 10 features in pre-trained model 2 with genetic data. Features were ranked according to descending order of absolute value of coefficient effect size.

From: Learning from Longitudinal Data in Electronic Health Record and Genetic Data to Improve Cardiovascular Event Prediction

Features

Reference gene

Coefficient

Age

—

0.747

rs17465637

MIA3

−0.334

rs67180937

MIA3

0.301

Gender

—

−0.270

EHR length

—

0.180

rs7568458

GGCX

0.103

rs4977574

CDKN2A

0.095

rs10455872

LPA

0.093

rs1412444

LPA

0.092

rs501120

CXCL12

−0.079

  1. We chose the result from the iteration which generated the closest result from the average AUROC and AUPRC.
Back to article page