Revealing disease-relevant alteration patterns through data aggregation

Abstract

It is important for both basic biological research and clinical applications that we not only understand the role of proteins as they are, but also how their function is affected by alterations in sequence and abundance. Towards this end, we present three contributions to the field addressing different aspects of this problem. Using clinical data and protein domain knowledge, we uncover two mutational patterns that affect hundreds of transcription factors in specific cancers, and analyze their significance using statistical techniques. Next, we introduce a statistical framework that utilizes paired gene expression profiles and histological images to identify proteins whose abundance associates with variation in morphological features in the tissue. Finally, we present a method to model the impact of amino acid substitutions using homolog alignments and experimental fitness measurements. We use this model to compute a directional substitution matrix based on experimental data and apply this matrix to new proteins for substitution impact prediction. Together, these methods and analyses serve to increase understanding of the downstream effects of protein alterations.