Multimodal imaging-based therapeutic fingerprints for optimizing personalized interventions: Application to neurodegeneration.

Personalized Medicine (PM) seeks to assist the patients according to their specific treatment needs and potential intervention responses. However, in the neurological context, this approach is limited by crucial methodological challenges, such as the requirement for an understanding of the causal disease mechanisms and the inability to predict the brain's response to therapeutic interventions. Here, we introduce and validate the concept of the personalized Therapeutic Intervention Fingerprint (pTIF), which predicts the effectiveness of potential interventions for controlling a patient's disease evolution. Each subject's pTIF can be inferred from multimodal longitudinal imaging (e.g. amyloid-ß, metabolic and tau PET; vascular, functional and structural MRI). We studied an aging population (N = 331) comprising cognitively normal and neurodegenerative patients, longitudinally scanned using six different neuroimaging modalities. We found that the resulting pTIF vastly outperforms cognitive and clinical evaluations on predicting individual variability in gene expression (GE) profiles. Furthermore, after regrouping the patients according to their predicted primary single-target interventions, we observed that these pTIF-based subgroups present distinctively altered molecular pathway signatures, supporting the across-population identification of dissimilar pathological stages, in active correspondence with different therapeutic needs. The results further evidence the imprecision of using broad clinical categories for understanding individual molecular alterations and selecting appropriate therapeutic needs. To our knowledge, this is the first study highlighting the direct link between multifactorial brain dynamics, predicted treatment responses, and molecular alterations at the patient level. Inspired by the principles of PM, the proposed pTIF framework is a promising step towards biomarker-driven assisted therapeutic interventions, with additional important implications for selective enrollment of patients in clinical trials.

Multifactorial causal model of brain (dis)organization and therapeutic intervention: Application to Alzheimer's disease.

Generative models focused on multifactorial causal mechanisms in brain disorders are scarce and generally based on limited data. Despite the biological importance of the multiple interacting processes, their effects remain poorly characterized from an integrative analytic perspective. Here, we propose a spatiotemporal multifactorial causal model (MCM) of brain (dis)organization and therapeutic intervention that accounts for local causal interactions, effects propagation via physical brain networks, cognitive alterations, and identification of optimum therapeutic interventions. In this article, we focus on describing the model and applying it at the population-based level for studying late onset Alzheimer's disease (LOAD). By interrelating six different neuroimaging modalities and cognitive measurements, this model accurately predicts spatiotemporal alterations in brain amyloid-ß (Aß) burden, glucose metabolism, vascular flow, resting state functional activity, structural properties, and cognitive integrity. The results suggest that a vascular dysregulation may be the most-likely initial pathologic event leading to LOAD. Nevertheless, they also suggest that LOAD it is not caused by a unique dominant biological factor (e.g. vascular or Aß) but by the complex interplay among multiple relevant direct interactions. Furthermore, using theoretical control analysis of the identified population-based multifactorial causal network, we show the crucial advantage of using combinatorial over single-target treatments, explain why one-target Aß based therapies might fail to improve clinical outcomes, and propose an efficiency ranking of possible LOAD interventions. Although still requiring further validation at the individual level, this work presents the first analytic framework for dynamic multifactorial brain (dis)organization that may explain both the pathologic evolution of progressive neurological disorders and operationalize the influence of multiple interventional strategies.