Clinical assessment of carotid artery disease currently relies on surrogate markers such as the degree of stenosis and patient symptomatology. However, a plethora of evidence indicates that carotid plaque instability is a multidimensional problem, where vulnerable plaques are identified by their internal composition, such as a large lipid-rich necrotic core and thin fibrous cap, rather than luminal narrowing alone. Despite the primary importance of plaque phenotype in determining the risk of complications such as ischemic stroke, there is a significant lack of individualized diagnostic tools and effective risk markers for carotid atherosclerosis. In this study we aim to design and optimize a network capable of predicting transcriptomic data (plaque bulk RNA-Seq) directly from non-invasive CTA imaging of plaque morphology. This foundational study involves the exploration of architectural and training variations to develop an optimized network capable of predicting plaque bulk RNA-Seq profiles from 3D CTA-imaged plaque morphology. By systematically refining the network's hyperparameters, the project aims to demonstrate that non-invasive imaging can be used to infer the underlying gene expression of a plaque, establishing the technical framework for subsequent multi-omics and spatial studies.