Table 1.

Summary of Top Performing v3 Models After Feature Selection Using RFE-CV

AA seq	Optimal Features (ap)	Residue Positions (Zm)	ROC AUC	Precision	Recall	F1	Acc
RbcL	101, 143, 281, 309, 418, 468	101, 143, 281, 309, 418, 468	1.0	1.0	0.9997	0.9998	0.9998
RpoC2	357, 368, 450, 774, 1050, 1132, 1290	357, 368, 450, 706, 942, 1011, 1151	0.9639	0.9728	0.9471	0.9591	0.955
NdhI	5, 25, 38, 49, 84, 88, 89, 147, 153	5, 25, 38, 49, 84, 88, 89, 147, 153	0.9583	0.9784	0.9197	0.9464	0.9423
NdhA	5, 29, 36, 98, 110, 187, 298, 301, 319, 320	5, 27, 34, 96, 108, 184, 293, 296, 314, 315	0.9805	0.9910	0.9031	0.9436	0.9401
RpoA	6, 14, 146, 163, 180, 243, 279, 326, 329, 336	6, 14, 146, 161, 176, 237, 270, 317, 327	0.969	0.9444	0.9305	0.9359	0.9295
MatK	49, 147, 159, 314, 378, 417, 436	16, 111, 123, 274, 338, 377, 396	0.9673	0.8491	0.9716	0.93	0.9191
NdhD	64, 76, 114, 334, 364, 376, 442, 451, 497, 501	62, 74, 112, 332, 362, 374, 440, 449, 495, 499	0.963	0.9214	0.9197	0.9183	0.9089
NdhF	89, 145, 287, 340, 400, 566, 568, 597, 659	89, 145, 287, 340, 400, 555, 557, 586, 646	0.9653	0.9860	0.8497	0.9114	0.9082

ap—positions numbered according to alignment position.

Z. m—Residue positions numbered according to positions in Zea mays protein sequence.

Average scores of classification metrics (ROC AUC, Precision, Recall, F1, Accuracy) after cross validation (using repeated random subsampling, n = 500, 70/30 train/test split) are shown. Gray cells indicate models and model performances that are potential artifacts of corresponding sequence length/variation. ROC AUC, area under receiver operating characteristic curve.