Table 2.

Predictive Model Results Describing the Performance of Random Forest Models at Predicting the Continuous Outcome Variables, (A) Birth Weight and (B) Placenta Weight

Molecular Signature Filter	Model Predictors a	Number of Molecular Predictors	Model Rank	RMSE Across Test Set	R2 of Regression Fit Between Predicted Versus Observed Values	RMSE of Regression Fit Between Predicted Versus Observed Values	% Variance Explained by Model
(A) Birth weight
Molecules meeting BH-adjusted p < .10	Covariates aloneb	0	—	161	0.27	57	23%
	mRNAs	254	3	141	0.45	71	41%
	miRNAs	268	2	137	0.46	69	44%
	CpG methylation sites	414	1	146	0.57	57	45%
Top 100 most significant molecules	Covariates aloneb	0	—	161	0.27	57	23%
	mRNAs	100	3	140	0.45	73	42%
	miRNAs	100	1	134	0.48	70	46%
	CpG Methylation Sites	100	2	148	0.52	65	44%
(B) Placenta weight
Molecules meeting BH-adjusted p < .10	Covariates alonec	0	–	124	0.41	46	35%
	mRNAs	11	2	135	0.32	54	29%
	miRNAs	70	1	128	0.36	48	31%
	CpG methylation sites	—	—	—	—	—	—
Top 100 Most Significant Molecules	Covariates alonec	0	—	124	0.41	46	35%
	mRNAs	100	3	155	0.08	63	7%
	miRNAs	100	1	126	0.40	43	33%
	CpG methylation sites	100	2	141	0.36	38	26%

Molecular signatures were filtered using a significance threshold of BH-adjusted p < .10 or top 100 ranking genes based on BH-adjusted p in association with each outcome. Models were ranked based on their performance across the prediction accuracy measures below, with 1 indicating the highest-ranking model and 3 indicating the lowest ranking model across molecular categories.

Abbreviation: RMSE, root mean squared error.

^aNote that all models included their corresponding covariates.

^bCovariates for birth weight included gestational age, exposure to first- or secondhand smoking, sex, and multiple gestations.

^cCovariates for placenta weight included BMI, multiple gestations, gestational age, and sex.