Pipeline design to identify key features and classify the chemotherapy response on lung cancer patients using large-scale genetic data

Table 9 Results of analysis of intersection of relevant SNPs given by the ML models, with GWAS Catalog records associated with LC and Cancer

Pipeline	# of	ML Rank	ML Rank	ML Rank
	features	cat ALL	cat LUNG	cat CANCER
RFE-LR + Up-sampling + RF	257	0	0	0
RLR-L1 + SMOTE-sampling + KNN	13	0	0	0
ANOVA + No sampling + RF	144	0	0	0
RFE-LR + SMOTE-sampling + RF	238	1	0	0
ANOVA + No sampling + Linear SVM	193	0	0	0
ANOVA + Up-sampling + Linear SVM	193	0	0	0
ANOVA + SMOTE-sampling + Linear SVM	193	0	0	0
RLR-L1 + SMOTE-sampling + RF	3	0	0	0
ANOVA + No sampling + KNN	95^a	0	0	0
RFE-LR + No sampling + RF	305	0	0	0
RFE-LR + No sampling + KNN	148^b	2	0	0
RLR-L1 + No sampling + KNN	17	0	0	0
RLR-L1 + Up-sampling + KNN	16	0	0	0
RFE-LR + Down-sampling + KNN	148^b	2	0	0
RFE-LR + No sampling + Linear SVM	148^b	2	0	0
RFE-LR + Up-sampling + Linear SVM	148^b	2	0	0
RFE-LR + SMOTE-sampling + Linear SVM	148^b	2	0	0
ANOVA + SMOTE-sampling + RF	193	0	0	0
RLR-L1 + No sampling + RF	17	0	0	0
ANOVA + Up-sampling + RF	193	0	0	0

^acorresponds to 5% of the top features selected by the ANOVA feature selection method. ^bcorresponds to 0,1% of the top features selected by the RFE-LR feature selection method

ISSN: 1752-0509