Table 2.

Performance metric summaries comparing training in withheld samples in PPMI.

Data Modality	Genetics (P < 1E-5)	Clinico-demographic	Transcriptomics (P < 1E-2)	Combined
Stage	Training in PPMI (70:30)	Training in PPMI (70:30)	Training in PPMI (70:30)	Training in PPMI (70:30)
Algorithm	MLPClassifier	LogisticRegression	SVC	AdaBoostClassifier
AUC (%)	70.66	87.52	79.73	89.72
Accuracy (%)	70.00	79.44	73.89	85.56
Balanced accuracy (%)	60.64	75.27	54.60	82.41
Log Loss	0.83	0.39	0.48	0.63
Sensitivity	0.83	0.85	0.97	0.89
Specificity	0.38	0.65	0.12	0.76
PPV	0.77	0.86	0.75	0.91
NPV	0.48	0.64	0.60	0.73