Disease risk prediction with rare and common variants

BMC Proceedings

Table 2 AUC values for models with and without rare variants in the second experiment

Variant selection p-value threshold	Common variants only	Common and rare variants	AUC value difference with 95% confidence interval
Size of training set = 300
0.01	0.7901	0.8442	0.0541 (0.0520, 0.0562)
0.005	0.8014	0.8364	0.0350 (0.0329, 0.0372)
0.001	0.8199	0.8437	0.0239 (0.0227, 0.0250)
5.0 × 10⁻⁴	0.8185	0.8429	0.0244 (0.0234, 0.0255)
1.0 × 10⁻⁴	0.8191	0.8333	0.0142 (0.0136, 0.0148)
5.0 × 10⁻⁵	0.8177	0.8310	0.0133 (0.0128, 0.0139)
1.0 × 10⁻⁵	0.8124	0.8279	0.0154 (0.0150, 0.0159)
Size of training set = 400
0.01	0.8315	0.8857	0.0542 (0.0524, 0.0561)
0.005	0.8348	0.8724	0.0376 (0.0362, 0.0391)
0.001	0.8346	0.8590	0.0244 (0.0236, 0.0253)
5.0 × 10^-4	0.8301	0.8564	0.0263 (0.0256, 0.0270)
1.0 × 10^-4	0.8247	0.8393	0.0146 (0.0143, 0.0150)
5.0 × 10^-5	0.8219	0.8355	0.0135 (0.0132, 0.0139)
1.0 × 10^-5	0.8157	0.8313	0.0156 (0.0152, 0.0159)
Size of training set = 500
0.01	0.8616	0.9182	0.0565 (0.0551, 0.0579)
0.005	0.8572	0.8966	0.0394 (0.0384, 0.0404)
0.001	0.8443	0.8686	0.0244 (0.0238, 0.0249)
5.0 × 10⁻⁴	0.8373	0.8636	0.0263 (0.0258, 0.0268)
1.0 × 10⁻⁴	0.8280	0.8422	0.0142 (0.0139, 0.0145)
5.0 × 10⁻⁵	0.8245	0.8378	0.0133 (0.0130, 0.0135)
1.0 × 10⁻⁵	0.8177	0.8331	0.0154 (0.0152, 0.0157)
Size of training set = 600
0.01	0.8850	0.9470	0.0619 (0.0610, 0.0629)
0.005	0.8730	0.9149	0.0418 (0.0411, 0.0425)
0.001	0.8510	0.8754	0.0244 (0.0240, 0.0248)
5.0 × 10⁻⁴	0.8420	0.8682	0.0262 (0.0259, 0.0265)
1.0 × 10⁻⁴	0.8303	0.8443	0.0140 (0.0138, 0.0141)
5.0 × 10⁻⁵	0.8259	0.8393	0.0133 (0.0132, 0.0135)
1.0 × 10⁻⁵	0.8188	0.8342	0.0155 (0.0154, 0.0156)

Prediction models were built on a randomly selected training data set for each replicate.

ISSN: 1753-6561