Annotation and function of switch-like genes in health and disease

Abstract

Gene expression microarrays provide transcript-level measurements across entire genomes and are traditionally used for differential expression analysis between health and disease or classification of disease subtypes. The abundance of gene expression microarray data currently available to the scientific community makes it possible to assess gene transcript levels among diverse tissue types for an entire genome. Gene expression is controlled over a wide range at the transcript level through complex interplay between DNA and regulatory proteins, resulting in gene expression profiles that can be represented as normal, graded, and bimodal (switch-like) distributions. It is our assertion that these distributions of gene expression, notably the bimodal distribution, result from biologically relevant regulation events. We have performed genome-scale identification and annotation of genes with bimodal, switch-like expression at the transcript level in human and mouse, using large microarray datasets for healthy tissue, in order to study the cellular pathways and regulatory mechanisms involving this class of genes. Our method implemented a likelihood ratio test to identify bimodal genes by comparing the best-fit two-component normal mixture, estimated using the expectation maximization algorithm, against a single-component normal distribution for each gene. This procedure identified roughly 15% of genes in human and mouse as bimodal, with a substantial overlap between human genes and their orthologous mouse counterparts. A survey of biological pathways revealed that the set of bimodal genes plays a role in cell communication and signaling with the external environment. Our analysis of regulatory sequence regions for bimodal genes revealed characteristics including enrichment of TATA boxes and an increased number of alternative transcription start sites. In addition to regulatory sequence analysis, we explored aspects of epigenetic regulation for their activity among the set of bimodal genes. We performed meta-analysis of gene expression microarray, DNA methylation, and histone methylation datasets representing human stem cells and liver tissue to reveal that the mode of expression within switch-like genes is primarily associated with histone methylation status. These results provide insight to normal patterns of histone methylation in healthy, differentiated tissue types. Aberrant methylation is a known marker in the progression of cancer, so these switch-like genes may also provide a valuable reference in disease diagnosis and prognosis. The method presented for bimodal gene identification also allows for an alternate approach to differential gene expression analysis between tissues and disease subtypes.%%%%Ph.D., Biomedical Engineering – Drexel University, 2008