Generalizable, high-throughput image analysis of subcellular structures using dispersion indices

Subcellular structural dynamics and relative spatial locations are fundamental for cellular function. However, quantification of subcellular changes relies on structural identifications that are labor-intensive, difficult to scale, and highly constrained to the specific biological system under investigation. We describe a computationally efficient and generalizable open-source algorithm GRID (generalized readout of image dispersion), which applies dispersion indices to quantify the diffusiveness and aggregation of subcellular structures based on the statistical properties of images. We demonstrate GRID in detecting changes in autophagic puncta, mitochondrial clustering, microtubule dynamics, and estimating half-maximal effective concentration (EC50) values in both 2D and 3D cell cultures. Using GRID, we further discover cell-type specific autophagy responses in multicellular human midbrain organoids. GRID enables high-throughput analysis of subcellular features that are critical for understanding disease mechanisms ranging from metabolic and neuronal diseases to cancer as well as a first-pass screening method for identifying biologically active agents for drug discovery.