Comparison and development of machine learning for thalidomide-induced peripheral neuropathy prediction of refractory Crohn’s disease in Chinese population

Table 1 Patients characteristics

Characteristics	Values (n = 164)
Male/Female	119/45
Age (yr)	34.3 ± 12.7
Thalidomide dosage mg/kg/d	1.5 (0.3-2.9)
White blood cell ( 109/L)	6.2 ± 2.7
Peripheral neuropathy	59 (36%)
Duration of thalidomide treatment (m)	17.2 (1-60)

Data are presented as median with range, mean ± SD or amount.

Table 2 Univariate analyses of factors associated with thalidomide-induced peripheral neuropathy

SNP	Gene	Minor allele	Main allele	OR	95%CI	P value
rs1353248	IL-12	T	C	8.983	2.497-30.90	0.0004
rs6265	BDNF	T	C	3.150	1.546-6.073	0.001
rs2030324	BDNF	G	A	3.164	1.561-6.434	0.001
rs11030104	BDNF	G	A	3.091	1.525-5.960	0.001
rs10991419	ABCA1	T	C	3.833	1.521-8.926	0.002
rs7795841	ABCB1	G	T	4.333	1.371-13.39	0.014
rs2575876	ABCA1	A	G	2.559	1.306-5.209	0.007
rs3918249	MMP9	T	C	3.800	1.208-10.50	0.016
rs7795846	ABCB1	A	G	3.690	1.155-10.25	0.020
rs62447181	IKZF1	A	G	2.933	0.763-9.514	0.096
rs11030100	BDNF	G	T	2.205	1.074-4.430	0.030
rs2777795	ABCA1	A	G	1.830	0.793-4.114	0.09
rs12718731	IKZF1	G	A	2.471	0.686-8.328	0.1
rs34165419	ABCA1	T	C	2.564	0.664-10.41	0.1

CI: Confidence interval; OR: Odds ratio; TiPN: Thalidomide-induced peripheral neuropathy; BDNF: Brain-derived neurotrophic factor; IL: Interleukin; SNP: Single nucleotide polymorphisms.

Table 3 Performance of the models for training set (all features)

Model	Precision	Sensitivity	Specificity	Accuracy	AUROC	F1 score
XGBoost	0.904	1	0.94	0.962	0.988	0.949
ET	0.526	1	0.471	0.667	0.999	0.69
GBDT	0.952	1	0.971	0.981	1	0.976
LR	0.613	0.95	0.647	0.759	0.907	0.745
RF	0.769	1	0.824	0.889	0.996	0.87

AUROC: Area under the receiver operating characteristic; XGBoost: Extreme gradient boosting; ET: Extremely random trees; GBDT: Gradient boosting decision tree; LR: Logistic regression; RF: Random forest.

Table 4 Performance of the models for testing set (all features)

Model	Precision	Sensitivity	Specificity	Accuracy	AUROC	F1 score
XGBoost	0.75	0.75	0.857	0.818	0.889	0.75
ET	0.667	1	0.667	0.8	0.833	0.8
GBDT	0.667	1	0.667	0.8	0.87	0.8
LR	0.8	0.667	0.889	0.8	0.741	0.727
RF	0.625	0.833	0.667	0.733	0.907	0.714

AUROC: Area under the receiver operating characteristic; XGBoost: Extreme gradient boosting; ET: Extremely random trees; GBDT: Gradient boosting decision tree; LR: Logistic regression; RF: Random forest.