Marker-free drift correction for 3D RNA imaging via spatiotemporal clustering and imputation

Spatially resolved RNA profiling is essential across various research fields. List-FISH, a recently developed RNA imaging technique, shows promising performance in high-resolution 3D RNA imaging of intact mouse brain sections. However, the application of List-FISH in centimeter-scale brain tissues is hindered by drift artifacts, which compromise the precision of FISH spot localization and the accuracy of cell recognition. Here, we introduce SCAM-Drift, a drift correction framework that integrates spatiotemporal clustering and missing value imputation for drift correction. Based on simulated data, fluorescent bead data, and synthetic data, we quantified the accuracy and robustness of SCAM-Drift for drift correction in List-FISH. Simulation results show that the correction precision is 65 nm for synthetic data with complete trajectories, and 216 nm for synthetic data with 50% trajectory loss. The test results from biological experiments show that SCAM-Drift achieved 23% improvement in cell recognition via corrected spots (quantified by ClusterMap). We believe that SCAM-Drift helps to promote the development and maturation of List-FISH for intact brain tissue sections.