Pan-cancer analysis of pre-diagnostic blood metabolite concentrations in the European Prospective Investigation into Cancer and Nutrition

Table 3 Robustness of the associations identified in the main analysis. For each identified association, the proportion of bootstrap samples on which it was replicated is reported (in bold when ≥50%)

Feature	Cancer type^a	Proportion of bootstrap samples^b	Proportion of bootstrap samples^c
Overall associations with cancer risk
c10	Overall	62%	59%
c4	Overall	47%	39%
Glutamine	Overall	73%	76%
Proline	Overall	65%	50%
lysopc_a_c18_2	Overall	57%	47%
pc_aa_c28_1_Clus	Overall	57%	64%
pc_aa_c32_2_Clus	Overall	49%	71%
pc_aa_c36_0_Clus	Overall	86%	95%
pc_aa_c36_1_Clus	Overall	50%	40%
Cancer type-specific associations
c14_1_Clus	BrC	80%	76%
Proline	BrC	77%	70%
pc_aa_c36_5_Clus	BrC	36%	47%
Arginine	CRC	88%	19%
his_Clus	CRC	81%	72%
pc_ae_c36_0	CRC	80%	46%
sm_c16_0_Clus	EnC	85%	87%
lysopc_a_c20_3_Clus	HCC	32%	47%
pc_aa_c40_2_Clus	HCC	61%	34%
sm_c16_0_Clus	HCC	90%	78%
c18_1_Clus	Adv.PrC	40%	49%
lysopc_a_c18_2	Loc.PrC	14%	23%
pc_aa_c36_0_Clus	Loc.PrC	49%	41%
h1	Loc.PrC	75%	68%

^aBrC stands for breast cancer, CRC for colorectal cancer, EnC for endometrial cancer, HCC for hepatocellular carcinoma, and Adv. Prc and Loc.PrC for advanced and localized prostate cancers, respectively
^bBootstrap samples were generated from the original sample of 5828 matched case-control pairs with information on 117 metabolites (corresponding to 50 features after the clustering step)
^cBootstrap samples were generated from the sample which comprised 4761 matched case-control pairs with information on 133 metabolites (corresponding to 65 features after the clustering step) after excluding the participants of the second CRC study

ISSN: 1741-7015