bioRxiv Science⌕ Search
Home/Archive/bioRxiv

bioRxiv · 10.1101/373878

Information-rich localization microscopy through machine learning

Kim, T.·Moon, S.·Xu, K.
View original source ↗Export metadata ↓

Abstract

While current single-molecule localization microscopy (SMLM) methods often rely on the target-specific alteration of the point spread function (PSF) to encode the multidimensional contents of single fluorophores, we argue that the details of the PSF in an unmodified microscope already contain rich, multidimensional information. We introduce a data-driven approach in which artificial neural networks (ANNs) are trained to make a direct link between an experimental PSF image and its underlying parameters. To demonstrate this concept in real systems, we decipher in fixed cells both the colors and the axial positions of single molecules in regular SMLM data.

Source connections

Explore related subjects

biophysicsPreprint

Keep this discovery

Explore connections, maps & timelines

BibTeXRIS

Kim, T., Moon, S., Xu, K.. 2018-07-22. Information-rich localization microscopy through machine learning. https://doi.org/10.1101/373878

Cite the original work for its findings. Save a collection to share your selection of sources.

KEEP EXPLORING

Related preprints

bioRxiv2026

Learning inhibitor interactions from protein deep mutational scanning data and deep learning models

Recent advances in protein language models, geometric deep learning, and structure prediction frameworks have the potential to accelerate computational drug discovery. However, it remains unclear how well these methods generalize beyond their training sets. Mutation-dependent drug responses present a challenge to these methods because there are often large effects from relatively small perturbations. Here, we systematically benchmark diverse mutational effect and drug-target affinity prediction models using deep mutational scanning datasets spanning 11 inhibitor conditions for the MET kinase domain. Across both mutational effect prediction and drug-target affinity tasks, current AI models produce largely ligand-invariant predictions despite experimentally distinct inhibitor-dependent fitness landscapes. Ligand-aware geometric models, including LigandMPNN and LASErMPNN, partially recover experimentally validated resistance-associated hotspots, highlighting the importance of explicitly modeling protein-ligand interactions and local structural flexibility. In contrast, deep learning affinity prediction models underestimate mutation-induced changes in binding affinity. This deficiency likely reflects mutation-invariant structural priors derived from co-folded structure prediction frameworks and the absence of mutation-conditioned supervision in current affinity datasets. Finally, we further extended our analysis to several kinase-inhibitor deep mutational scanning datasets and confirm the overall trends. Our findings suggest that a primary limitation in improving performance on this task lies in the scarcity of mutation-conditioned protein-ligand training data rather than model architecture alone.

biophysics↗
bioRxiv2026

Particle picking and centering in cryo-electron micrographs via a vanishing radial kernel

A training-free procedure combines particle picking and approximate centering in cryo-electron microscopy through a single convolution of the micrograph with a compact radial kernel, whose weights increase with distance from the origin, saturate in an annulus, and vanish beyond it. Particles are darker than their surroundings and much weaker than the noise; each carves a well in the response, and local minima of the response are candidate centers. Constant background offsets leave these extrema unchanged away from image boundaries, and a local stability bound controls the displacement of a well's minimum under signed perturbations without assuming a symmetric particle. Grid seeds are iteratively updated, nearby endpoints are merged, and two-dimensional classification follows. On the \emph{Plasmodium falciparum} 80S ribosome and $\beta$-galactosidase data sets, reconstructions reach 3.68 and 4.30~\AA{}, compared with 3.77 and 4.36~\AA{} using reference coordinates, with picking at 0.20--0.26~s per micrograph.

biophysics↗
bioRxiv2026

Comparative cryo-EM structures of CGMMV and ToMMV reveal divergent coat protein packing and electrostatics indicating Tobamovirus host specificity

Tobamoviruses are a group of economically important plant viruses exhibiting extreme stability and persistence. Despite high morphological similarity, they show extensive host-family specificity, cause a range of symptoms, and cause significant global yield loss in fruit and vegetable plants. Composed of a rod-shaped helical assembly of identical coat proteins encasing a single-stranded RNA, Tobamoviruses have long served as fundamental models for understanding helical assemblies. Here we report the cryo-EM structures of two viruses from this genus, CGMMV and ToMMV, at 2.86 [A] and 2.71 [A] resolution, respectively. The structures reveal divergent packing strategies within the virion assembly. While ToMMV is genetically similar to the canonical Tobamovirus TMV and displays a structure closely resembling TMV, CGMMV employs a more hydrophobic packing strategy, particularly in the solvent-exposed inner loop region of the coat protein. Comparative biophysical and molecular dynamics analyses from the reconstructed structures reveal more stable and compact protein packing for CGMMV at both the subunit and assembly levels. Furthermore, this study reports a newly identified electrostatic patch existing at the lower alpha-helix region of the coat protein that serves as an additional RNA-stabilizing region, with host family-specific amino acid substitutions among Tobamoviruses. Our study elucidates the structural mechanics and evolutionary strategies driving virion stability and host diversity while maintaining a conserved helical assembly. These findings reveal how subtle electrostatic tuning and divergent packing strategies drive viral stability and host adaptation within a strictly conserved quaternary scaffold.

biophysics↗