Actin Filament Tracing

Summary: BundleTrac is an effective method for tracing hundreds of filaments in an Actin bundle. It is a semi-automatic modeling approach in which a seed point is provided for each filament. BundleTrac works in two steps. The first step improves the filamentous pattern present in the image using cross-correlation-based averaging. In the subsequent step, individual filaments are traced employing a 7-peaks Gaussian kernel and 2D-convolution optimization. When applied to an experimental density image of Stereocilium, BundleTrac yields high agreement with manual annotation

Summary: Spaghetti Tracer is  a dynamic programming-based automatic framework designed to characterize the structural arrangement of semi-regular filaments within complex 3D maps of subcellular components. Assuming the tomogram can be rotated to align filaments along a mean direction, Spaghetti Tracer initially identifies local seed points for potential filament segments. These segments are then expanded from the seeds using a dynamic programming algorithm. To assess its effectiveness, we validate various algorithmic variations of our framework using simulated tomograms that closely resemble the noise and appearance of real experimental maps. By leveraging the known ground truth within these simulations, we conduct a statistical analysis, including precision, recall, and F1 scores, which demonstrates  the efficacy of Spaghetti Tracer.

Summary: Struuwel Tracer is a novel computational framework for tracing randomly-oriented actin filaments in cryo-electron tomography maps. The method accumulates densities along paths from seed points, forming short linear filament segments that are then scrutinized and classified using a pruning map. These segments are iteratively fused into longer and curved filaments based on various criteria. Testing on simulated tomograms of Dictyostelium discoideum filopodia and experimental tomograms of mouse fibroblast lamellipodia shows high efficacy, with F1-scores ranging from 0.85 to 0.90.