Project Summary: The parent award (R01MH125615) of this supplement application seeks to investigate the neural mechanisms of learning and un-learning of a fear response through fear conditioning and fear extinction. The goal is to advance our understanding of how visual and attention networks interact during associative learning as well as to inform clinical intervention and diagnostic procedures in a variety of psychiatric disorders where fear is a transdiagnostic pathology. A large multimodal/multiscale neuroimaging dataset, which includes simultaneous EEG-fMRI, physiological measures such as heart rate and skin conductance, as well as behavioral and self-report data, is being acquired according to the proposed timeline. Recent advances in AI/ML are beginning to revolutionize neuroimaging and neural data analysis. We seek to leverage these advances to enable innovative testing of our hypotheses. Readying our multimodal/ multiscale data for AI/ML, however, faces challenges. The goal of this administrative supplement is to bring together expertise in data management, data processing, AI/ML, and neuroscience/experimental psychology to meet these challenges. Two aims will be pursued. The objective of Aim 1 is to build an infrastructure for preparing the multimodal/multiscale data for AI/ML analysis and sharing. Specifically, we will develop protocols for data denoising, imputation, pre-processing, bias correction, artifact removal, normalization, and harmonization and establish pipelines to integrate and consolidate data from different data sources into a unifying repository for analysis and sharing according to the FAIR principle. The objective of Aim 2 is to design a novel AI-driven platform for analyzing multimodal/multiscale data. Specifically, we will develop a transformer-based platform to enable multimodal learning from diverse sources of data and link the outcomes of learning with the proposed cognitive/neurophysiological model to enable the innovative testing of the hypotheses in the parent award.