From Bayesian modeling to genomic mapping: Biologic validity of predictive single nucleotide polymorphism networks for chemotherapy-related side effects.

Abstract

9535 Background: Using proprietary Bayesian network (BN) algorithms, we discovered interacting single nucleotide polymorphism (SNP)-BNs predicting patient-risk for 6 chemotherapy (CT)-induced side effects (SEs): oral mucositis (OM), diarrhea (D), CT-induced nausea and vomiting (CINV), fatigue, cognitive dysfunction (CD) and peripheral neuropathy (PN). To assess biologic validity, SNP-BNs mapped each SE to its gene loci, exploring biologic pathways for those genes. Roles of genes and pathways found in similar phenotypic diseases (e.g., Inflammatory Bowel Disease [IBD] for D, Alzheimer disease [AD] for CD, chronic fatigue syndrome [CFS] for F) were explored. Methods: Via an FDA-cleared DNA extraction kit, saliva was collected from 78 women with breast cancer (dose-dense AC+T) and 57 patients with colon cancer (FOLFOX +/- bevacizumab). We assessed 2.5M SNPs/patient (Illumina OmniQuad bead chip assays). SEs were measured via a validated tool (Patient Care Monitor). Analyzed SNP arrays and SEs discovered SNP-BNs. Final SNP-BNs per SE included gene name, symbol, and loci. Pathways and networks were analyzed for gene function and related genes, querying 6 databases (NLM, NextBio, Weizmann Institute, SPRING, Gene Ontology, AmiGO). Results: Mucosal injury-implicated networks and genes were identified, including oxidative stress, NF-κB, pro-inflammatory cytokines IL-1, IL-6, and TNF, MMPs and fibronectin breakdown (SPOCK3). D networks and genes overlapped with IBD (ERG, IER2, CASK). SNP-BNs for CINV overlapped in both CT regimens with genes for neurotransmitters (serotonin, dopamine, substance P) and nerve impulse transmission. B3GAT1, involved with opioid-induced CINV, was identified. F networks involving genes and pathways for cytokines and skeletal muscle overlapped to CFS. CD networks involve genes in AD. PN networks mapped genes involving myelination, demyelination, inflammation, and toxic neuropathy. Conclusions: SNP-BNs predicting CT-induced SEs mapped to genes associated with known biology of each SE, sharing central traits with other diseases, and are targets for drug discovery in cancer-related SEs and unmet needs, including AD.