Biomarker-guided translation of brain imaging into disease pathway models

Abstract

The advent of state-of-the-art brain imaging technologies in recent years and the ability of such technologies to provide high-resolution information at both structural and functional levels has spawned large efforts to introduce novel non-invasive imaging biomarkers for early prediction and diagnosis of brain disorders; however, their utility in both clinic and drug development at their best resolution remains limited to visualizing and monitoring disease progression. Given the fact that efficient translation of valuable information embedded in brain scans into clinical application is of paramount scientific and public health importance, a strategy is needed to bridge the current gap between imaging and molecular biology, particularly in neurodegenerative diseases. As an attempt to address this issue, we present a novel computational method to link readouts of imaging biomarkers to their underlying molecular pathways with the aim of guiding clinical diagnosis, prognosis and even target identification in drug discovery for Alzheimer's disease.

Figure 1. Generation of brain region-specific subnetwork models using imaging readouts.

The proposed methodology - in the first step - incorporates information of diagnosed brain regions from imaging into the brain interactome annotated with 15 brain regions.

Figure 2. Enhancement of subnetwork models with the information of potential biomarkers.

Enhanced subnetwork models are further subjected to pathway analysis guided by biomarker pins.

Figure 3. Representation of recovered HIV-NEF pathway and its first neighbor proteins in the temporal lobe network model.

The model has been enhanced with drug-target and biomarker expression information. Circular nodes in yellow show membership to the HIV-NEF pathway; square nodes in pink are approved drugs targeting the recovered pathway; triangle nodes represent potential biomarkers color coded for their expression levels in AD brain (red: over-expression; green: under-expression).

Figure 4. The extended algorithm of the core methodology for translation of imaging information.

The proposed image translation methodology can be further extended to an algorithm, which could be potentially used for clinical decision-making support.