Extracting biological structure and heterogeneity from the nano to the macro scale
Rosten, E.; Stedman, D.; Chu, L.-Y.; Wickramanayake, K.; Littlejohn, G.; Baxter, K. J.; McConnell, G.; Culley, S.; Ch'ng, Q.; Leterrier, C.; Bates, M.; Molodtsov, M.; Cox, S.
Show abstract
Fluorescence microscopy is an essential tool in biology. It has revealed great variability at multiple scales, in macromolecular complexes, cells, and organisms. Understanding this variability will reveal the mechanisms by which genetically or biochemically identical systems adopt different biological states. Achieving this requires the ability to extract both the underlying biological structure and how it varies across the population. Currently the field lacks general techniques to deal with arbitrary structures and different types of variability. Here we present SQUASSH, a new convolutional neural network-based approach to freely fit structural models to fluorescence microscopy data that simultaneously quantifies variability to reveal correlations, dynamics, and systematic distortions. SQUASSH is highly versatile: it accommodates diverse imaging modalities at length scales from nm to mm. This approach opens up applications such as imaging nanoscale macromolecular structures, revealing patterns in shape changes from organelle to tissue scale, and characterizing systems biology of dynamical processes.
Matching journals
The top 4 journals account for 50% of the predicted probability mass.
Similar papers in this journal
- Deep neural network automated segmentation of cellular structures in volume electron microscopy 96%
- CLEMSite, a software for automated phenotypic screens using light microscopy and FIB-SEM 93%
- Implicit Laplacian of Enhanced Edge: An Unguided Algorithm for Accurate and Automated Quantitative Analysis of Cytoskeletal Images 92%
Similar papers in this journal
"Similar papers" are the closest papers from that journal in the model's embedding space. They show what the match is built on, but the ranking comes mostly from a classifier over the whole training set, not from these examples alone.