Table 7 The best candidate models of interactions of selected attributes related to lung cancer by MDR approach.

Model factor^a,b	Attribute	Bal. Acc.^c overall	Bal. Acc. CV^d training	Bal. Acc. CV testing	CV consistency	P value^e
Whole group
Lung cancer
IL1B + PPP1R13L + POLR1G + smoking-duration
One-factor	Smoking	0.5708	0.5708	0.5708	10/10	0.0040–0.0050
Two-factor	Smoking
Two-factor	rs1143633	0.5849	0.5866	0.5649	5/10	0.0090–0.0100
Three-factor	Smoking
	rs3136558
	rs967591	0.6062	0.6084	0.5602	8/10	0.0100–0.0110
IL1B + smoking-duration
One-factor	Smoking	0.5708	0.5708	0.5708	10/10	< 0.0010
Two-factor	Smoking
Two-factor	rs1143633	0.5849	0.5862	0.5721	6/10	< 0.0010
Three-factor	Smoking
	rs1143633
	rs1143630	0.5944	0.5964	0.5619	8/10	< 0.0010
Fourth-factor	Smoking
	rs1143633
	rs3136558
	rs1143630	0.6112	0.6139	0.5588	10/10	0.0040–0.0050
IL1B + PPP1R13L + POLR1G
Two-factor	rs1143633
Two-factor	rs1970764	0.5682	0.5689	0.5507	9/10	0.0440–0.0450
Three-factor	rs1143633
	rs1970764
	rs735482	0.5973	0.5987	0.5576	10/10	0.0130–0.0140
IL1B
One-factor	rs1143633	0.5451	0.5451	0.5451	10/10	0.0290–0.0300
Histopahtology subgroup
Lung squamous cell carcinoma
IL1B + smoking-duration
One-factor	Smoking	0.6315	0.6318	0.6227	10/10	< 0.0010
Two-factor	Smoking
Two-factor	rs1143633	0.6532	0.6533	0.6472	10/10	< 0.0010
Three-factor	Smoking
	rs1143633
	rs1143630	0.6619	0.6645	0.6123	5/10	< 0.0010
Fourth-factor	Smoking
	rs1143633
	rs3136558
	rs1143630	0.6785	0.6827	0.5871	10/10	0.0030–0.0040
Other
IL1B + smoking-duration
Two-factor	Smoking
Two-factor	rs3136558	0.6178	0.6181	0.6073	10/10	0.0090–0.0100

^aAnalyzed by MDR 3.0.3. dev. Jar, data of PPP1R13L + CD3EAP from previous study (Yin et al.¹⁸).
^bOnly list statistical significant models.
^cBalanced accuracy.
^dCross-validation.
^eP value based on 1000 permutation test. Boldface indicates statistical significance.

Search