Classification of Diffuse Subcellular Morphologies

• DOI: 10.25080/majora-1b6fd038-00f
• Publication Date: 2021-08-01T19:34:17Z
• Authors: Neelima Pulagam; Marcus Hill; Mojtaba Fazli; Rachel Mattson; Meekail Zain; Andrew Durden; Frederick Quinn; S Chennubhotla; Shannon Quinn
• Keywords: Organelle

Characterizing dynamic sub-cellular morphologies in response to perturbation remains a challenging and important problem. Many organelles are anisotropic difficult segment, few methods exist for quantifying the shape, size, quantity of these organelles. The OrNet (Organelle Networks) framework models diffuse organelle structures as social networks using graph theoretic probabilistic approaches. Specifically, this architecture tracks morphological changes mitochondria because its structural offer insight into adverse effects pathogens on host aid diagnosis treatment diseases; such tuberculosis. offers segmentation pipeline preprocess confocal imaging videos that display various mitochondrial network graphs. Earlier anomaly detection include manual identification by researchers biology domain. Although those approaches were successful, classification is time consuming, tedious, error-prone. Existing convolutional architectures do not have capability adapt general graphs fail represent due their amorphous characteristic. Thus, we propose two different perform captures behaviors identifies fragmentation fusion mitochondria. One deep learning architecture, second an approach finds representation each uses traditional machine method classification. Recent studies demonstrated neural well time-series tasks, better able spatially Alternatively, much research has established be promising robust models. Testing comparing will effectively improve robustness categorizing distinct subcellular very useful identifying infection patterns, offering new way understand cellular health responses.