Table 1 Ranking of top gene features based on SHAP analysis, showing mean absolute SHAP values, normalized contributions, and cumulative proportions used to define the prioritized biomarker set.

From: Explainable machine learning identifies immune-inflammatory biomarkers and therapeutic candidates in drug-resistant epilepsy

Gene

MeanAbsSHAP

NormMeanAbsSHAP

Cumulative

XIRP1

0.012882

0.03727

0.03727

P2RY13

0.010481

0.030325

0.067595

CCL3-AS1

0.008491

0.024566

0.092161

ACY3

0.007691

0.022253

0.114414

TNF

0.007405

0.021425

0.135839

DHRS9

0.006998

0.020248

0.156087

CX3CR1

0.006825

0.019747

0.175834

CCL3

0.006159

0.017821

0.193655

CD163

0.006098

0.017643

0.211298

PLD4

0.005978

0.017297

0.228595

P2RY12

0.0058

0.016781

0.245375

S100A9

0.005691

0.016466

0.261841

SELPLG

0.005614

0.016242

0.278084

MIR17HG

0.004945

0.014306

0.29239

S100A8

0.003859

0.011164

0.303554

IL1A

0.003383

0.009787

0.313341

CCL4

0.003341

0.009666

0.323007

IL1B

0.003319

0.009603

0.33261

SYN3-AS1

0.003256

0.009421

0.342031

B3GNT5

0.003141

0.009088

0.351119