Pathway Analysis of Genome-wide Association Study in Childhood Leukemia among Hispanics

Abstract

Background The incidence of acute lymphoblastic leukemia (ALL) is nearly 20% higher among Hispanics than non-Hispanic Whites. Previous studies have shown evidence for association between risk of ALL and variation within IKZF1, ARID5B, CEBPE, CDKN2A, GATA3, and BM1-PIP4K2A genes. However, variants identified only account for <10% of the genetic risk of ALL. Methods We applied pathway-based analyses to GWAS data from the California Childhood Leukemia Study to determine whether different biological pathways were overrepresented in childhood ALL and major ALL subtypes. Furthermore, we applied causal inference and data reduction methods to prioritize candidate genes within each identified overrepresented pathway, while accounting for correlation among SNPs. Results Pathway analysis results indicate that different ALL subtypes may involve distinct biological mechanisms. Focal adhesion is a shared mechanism across the different disease subtypes. For ALL, the top five overrepresented Kyoto Encyclopedia of Genes and Genomes pathways include axon guidance, protein digestion and absorption, melanogenesis, leukocyte transendothelial migration, and focal adhesion (PFDR < 0.05). Notably, these pathways are connected to downstream MAPK or Wnt signaling pathways which have been linked to B-cell malignancies. Several candidate genes for ALL, such as COL6A6 and COL5A1 were identified through targeted maximum likelihood estimation. Conclusions This is the first study to show distinct biological pathways are overrepresented in different ALL subtypes using pathway-based approaches, and identified potential gene candidates using causal inference methods. Impact The findings demonstrate that newly developed bioinformatics tools and causal inference methods can provide insights to furthering our understanding of the pathogenesis of leukemia.

Novel Gene and Network Associations Found for Acute Lymphoblastic Leukemia Using Case–Control and Family-Based Studies in Multiethnic Populations

“…This subnetwork shows genes such as CEBPE , which contains genome-wide significant SNPs, interacting with other genes involved in hematopoiesis. These networks have not been identified in previous pathway analyses on ALL (24). …”

“…We also find that a significant subnetwork centered on MEIS1 is only identified in network analysis of the Hispanic American cohort-derived PEGASUS gene scores, and is missed in network analysis of the European American cohort (Figure 2B). These candidate genes were not identified in a previous analysis of pathways in Hispanic individuals and can represent useful targets for future functional validation (24). …”

“…The goal of our study is similar to that of Hsu et al (24) -- to identify genes and gene sets associated with ALL risk -- but our approaches are quite distinct. First, PEGASUS (20) reports a gene score for each gene in the genome that is sensitive to genes containing multiple variants of moderate association with a trait of interest while controlling for LD, which may vary with ancestral background.…”

“…PEGASUS gene scores are not limited by pre-existing annotations and allow for the calculation of false discovery rates. Hsu et al (24) do not calculate gene scores, but instead use a GWA SNP-level threshold ( p < 0.001) to identify candidate genes for gene-set enrichment analysis. We also note that the pathways that Hsu et al (24) identify do not contain a number of known ALL-associated genes, whereas our network and pathway results (Figures 1 and 2, Table 2) contain genes such as CEPBE and IKZF1 , which have been identified previously in GWA studies of ALL.…”

“…Recent work from our group and others has yielded novel methods for exploring gene networks in the context of a GWA framework (16–23), including an analysis of gene sets associated with ALL by Hsu et al (24), which is discussed in detail later. In the gene-level method, PEGASUS (20), association p -values are combined within genes while correcting for linkage disequilibrium (LD); this approach allows us to account for genetic heterogeneity -- when different causal mutations of small effect in the same gene or pathway may be present across cases -- and to test for genes and pathways significantly associated with ALL.…”