Uncovering the dark transcriptome in polarized neuronal compartments with mcDETECT

Abstract

Spatial transcriptomics (ST) is a powerful tool for studying the molecular basis of brain diseases. However, most current analyses focus only on nuclear or somatic expression, overlooking distally localized mRNAs, known as the "dark transcriptome." These transcripts, which make up nearly 40% of the brain transcriptome, are packaged into RNA granules and transported to polarized neuronal structures critical for neuronal function. Here, we present mcDETECT, a machine learning framework that leverages in situ ST (iST) data to detect RNA granules and resolve their RNA composition. Applying mcDETECT to simulated and real datasets from multiple iST platforms, we demonstrate robust granule detection and molecular subtyping. Granule-informed downstream analysis further revealed functionally distinct neuronal substates not captured by somatic expression alone. In an Alzheimers disease (AD) mouse model, mcDETECT uncovered alterations in RNA granule distributions and associated neuronal substates prior to measurable neuronal loss, highlighting potential therapeutic targets for early AD pathology.