In the second part of this series, I review four “Brief communicaions” papers.
Unsupervised modeling of cell morphology dynamics for time-lapse microscopy
This work, from the lab of Daniel Gerlich, provides a tool for automatic prediction and annotation of cell morphologies from time-lapse microscopy experiments.
Their initial task was to annotate the six different cell-cycle mitotic phases of human HeLa cells which express histone 2B tagged with mCherry. Tagging H2B enabled us to visualize the chromosomes. Following time-lapse microscopy, they let three biologists to annotate the images – determine the stage of the cell in each image. They found mild inconsistencies in the annotations of the same person on different days, and an extensive variation between the three biologists.
This is not surprising. Although scientists are supposed to be as objective and impartial as can be, sometimes different people notice different things in the same image.
Therefore, the authors thought it is a good idea to develop a program for objective data annotation in an automated fashion. I am not qualified to judge the quality of their work. I could not even follow some of the terminology. However, the authors claim that their program can objectively annotate cell morphologies in a time lapse experiment (i.e. changes in the same cell over time) quite successfully.
The problem with this program, if I understand correctly (I’m not sure I am), is that you need to give the program some standard base-line annotation. In this paper, they used a “majority vote” for each stage to achieve a sort of consensus base-line for the program. But once you get that baseline, the program does better than individual researchers. In their assays, it could successfully detect deviations from the time cell-cycle when they used RNAi to affect cell-cycle factors. That’s very nice.
An image analysis toolbox for high-throughput C. elegans assays
C. elegans is a useful organism to study many biological processes. There are many genetic and other tools that are used in worm research, particularly in large-scale screens. One of the first phenotypes that researchers look at is worm morphology. However, such morphological distinctions are still done manually, in most cases. The lab of Anne Carpenter developed an open-source toolbox to identify different morphologies, both from bright-field images and fluorescent images. Worm scientists will love it!
Fluorescently tagged worms. (a) worms in their natural form. (b) All the worms in (a) are juxtaposed after automated detection and digital straightening. (c) A low-resolution worm atlas enables feature measurements in relation to the worm’s anatomy. Source: Wahlby et al. Nature Methods 9,714–716 (2012)
Elastic volume reconstruction from series of ultra-thin microscopy sections
The lab of Pavel Tomancak developed a program that creates 3D reconstructions from electron microscopy serial images. I did not understand the math and programming terminology, but their videos are cool.
Faster STORM using compressed sensing
STORM, STochastic Optical Reconstruction Microscopy, aka Photoactivated localization microscopy (PALM), is a super resolution microscopy method which is based on single molecule stochastic switching. What does it mean? Most fluorophores “active” all the time, meaning that they emit light in response to the excitation light. However, several photoactivatable fluorophores have been developed. This means that the molecules are dark even when shined upon by the excitation light. Only after activating the molecules by shining a certain wavelength (usually UV light, but not exclusively), the molecules are emit light. This is a huge advantage in fluorescent microscopy, since it reduces the background level; it allows activation of the fluorophore only at a certain region of interest; it allows very accurate time-resolution experiments and is very usefull for super-resolution experiments. This paper, from the lab of Bo Huang, deals with a technical problem in super-resolution microcopy which uses STORM. Apparently, to achieve super-resolution, one needs to take thousands of frames or acquire for tens of seconds. Thus, the temporal resolution is pretty low. The temporal resolution can be improved by increasing the excitation intensity (but risking photodamage to the camera and/or fluorophores) or increasing the density of the fluorophore (thus risking them to overlap, invalidating to common single-molecule algorithms). There are other algorithms, but if I understand their claims, these algorithms require an estimation of the number of molecules in the image. Here’ they describe novel approach, which they claim can get them to a three second time resolution.
To be continued….
Zhong Q, Busetto AG, Fededa JP, Buhmann JM, & Gerlich DW (2012). Unsupervised modeling of cell morphology dynamics for time-lapse microscopy. Nature methods PMID: 22635062
Wählby C, Kamentsky L, Liu ZH, Riklin-Raviv T, Conery AL, O’Rourke EJ, Sokolnicki KL, Visvikis O, Ljosa V, Irazoqui JE, Golland P, Ruvkun G, Ausubel FM, & Carpenter AE (2012). An image analysis toolbox for high-throughput C. elegans assays. Nature methods, 9 (7), 714-716 PMID: 22522656
Saalfeld S, Fetter R, Cardona A, & Tomancak P (2012). Elastic volume reconstruction from series of ultra-thin microscopy sections. Nature methods, 9 (7), 717-720 PMID: 22688414
Zhu L, Zhang W, Elnatan D, & Huang B (2012). Faster STORM using compressed sensing. Nature methods, 9 (7), 721-723 PMID: 22522657