Combining epidemiological and in silico investigations to elucidate the genetic basis of the inverse occurrence between Alzheimer's disease and cancer

The inverse relationship of occurrence of Alzheimer's disease (AD) and cancer has been reported in several population based studies and meta-analysis. The different distribution of exposures to known risk factors in people with cancer or AD was found insufficient to explain, in our previous cohort study, the lower than expected co-occurrence. Aim of this study was to investigate the genetic basis underlying the inverse relationship by analysing common genetic variants differently expressed in the two diseases and their potential role in molecular pathways. A comparative analysis explored the distribution of 545,982 known Single Nucleotide Polymorphisms (SNPs) in the genome of large populations of cancers (N=4409) and AD (N=1292) patients from NIH dbGaP datasets. GWAS, PCA methods, SNPnexus, STRING, and ToppGene tools were employed to compare the datasets and distinguish molecular processes associated with AD and cancer. The GWAS analyses identified 300 SNPs (p< 10-5) associated with 213 unique genes (SNPnexus functional annotation) lying within 1 Mb from each SNPs' position. The gene set enrichment analysis (GSE) in ToppGene identified 11 out of 213 genes as significantly (p<10-5) enriching phospholipid binding (GO:0005543): ABCA1, CADPS, GBF1, KCNQ1, MARCKS, NF1, PLD1, PXK, SNX29, TIAM1, ZFYVE26. The protein-protein interaction analysis (STRING) and exploration of activity of proteins produced by the 11 genes, indicated enrichment for molecular pathways associated to carcinogenesis (RAS) and apoptosis. Furthermore we tested the SNPs ability of discriminating AD form cancer by calculating a composite numeric score based on the 40 most significant ones (p<10-12) from the GWAS analysis. For each of these SNPs we assigned value 1 if an allele with minor frequency was observed, and 0 otherwise. In a Receiver Operating Characteristic analysis the score performance resulted high, with AUC=83.1 [95% CI 82.0-84.2]. The adopted combination of epidemiological and in silico approaches indicates potential explanatory capabilities. The biological soundness of our finding is consistent with the existing studies identifying the cell membrane damage and metabolic processes as underlying initiation and progression of both AD and cancer. These preliminary results show that genes, identified by SNPs significantly different in the two diseases, are involved in shared biological pathways that, if deregulated, may explain the divergent trajectories towards AD or cancer. Further investigations using other independent genetic datasets are required to confirm these findings that, if successfully replicated in silico, can form the basis for specific in vitro and in vivo future studies on the inverse occurrence of the two diseases.