Graduation Date

Fall 12-16-2016

Document Type


Degree Name

Doctor of Philosophy (PhD)


Biomedical Informatics

First Advisor

Chittibabu (Babu) Guda


The human brain is the most complex structure known to mankind and one of the greatest challenges in modern biology is to understand how it is built and organized. The power of the brain arises from its variety of cells and structures, and ultimately where and when different genes are switched on and off throughout the brain tissue. In other words, brain function depends on the precise regulation of gene expression in its sub-anatomical structures. But, our understanding of the complexity and dynamics of the transcriptome of the human brain is still incomplete. To fill in the need, we designed a gene expression model that accurately defines the consistent blueprint of the brain transcriptome; thereby, identifying the core brain specific transcriptional processes conserved across individuals. Functionally characterizing this model would provide profound insights into the transcriptional landscape, biological pathways and the expression distribution of neurotransmitter systems.

Here, in this dissertation we developed an expression model by capturing the similarly expressed gene patterns across congruently annotated brain structures in six individual brains by using data from the Allen Brain Atlas (ABA). We found that 84% of genes are expressed in at least one of the 190 brain structures. By employing hierarchical clustering we were able to show that distinct structures of a bigger brain region can cluster together while still retaining their expression identity. Further, weighted correlation network analysis identified 19 robust modules of coexpressing genes in the brain that demonstrated a wide range of functional associations. Since signatures of local phenomena can be masked by larger signatures, we performed local analysis on each distinct brain structure. Pathway and gene ontology enrichment analysis on these structures showed, striking enrichment for brain region specific processes. Besides, we also mapped the structural distribution of the gene expression profiles of genes associated with major neurotransmission systems in the human. We also postulated the utility of healthy brain tissue gene expression to predict potential genes involved in a neurological disorder, in the absence of data from diseased tissues. To this end, we developed a supervised classification model, which achieved an accuracy of 84% and an AUC (Area Under the Curve) of 0.81 from ROC plots, for predicting autism-implicated genes using the healthy expression model as the baseline. This study represents the first use of healthy brain gene expression to predict the scope of genes in autism implication and this generic methodology can be applied to predict genes involved in other neurological disorders.