The power of Deep Neural networks

I am wowed at the images published recently in BioArxiv in the Zero-shot deconvolution networks paper. It will be interesting to see the peer-reviewed paper.

I would be interested in visualizing the amount of errors made by the network. A simple way to do it is to acquire a short time lapse (eg 10 images) of a fixed sample, run it through the network and see which structures are stably identified and which change from frame to frame. 🙂

 

Light-seq: Multiplexed, non-destructive spatial transcriptomics of tissues sections using light

Light-Seq is a new pretty cool technique for highly multiplexed sequencing of RNA in tissue sections using light. This technique is highly sensitive, highly spatially resolved and because it does not destroy the tissue, it can be combined with protein labelling (genetic or by immunolabelling).

On one of our single-point confocal/spinning disk/widefield system at the LCI facility, we have a device called Primo (DMD + UV laser) which can be used to run this technique! 🙂

Let us know if you would like to set up LightSeq at the LCI core facility!

mCherry-XL: brigther, more stable and with better spectra!

mCherry is a very popular red fluorescent protein. However it has several disadvantages:

  • It is shifted towards far red (ex peak 585 nm) so it often is not imaged optimally with the illumination sources and filters commonly available.
  • It bleaches fast

mCherry has now been evolved into mCherry-XL with several improvements:

  • This variant is shifted back towards green (ex peak 560 nm) therefore being very well excited with popular 561nm lasers.
  • It is 3 times brighter than mCherry
  • There is also a clear improvement in the lifetime for FLIM
  • Together the 2 points above means that less excitation power is required so it should help with the bleaching problem

Here is the paper. Therefore you should consider mCherry-XL for your future tagging with red fluorescence proteins.

CLIJ2: Open Computing Language and ImageJ2, GPU power for everyone

CLIJ2 allows you to use ImageJ/Fiji on GPU instead if CPU processing, so much faster! 🙂

Here is a nice article about what CLIJ2 can do. By the way, this article is published on a new imaging forum called FocalPlane. Check it out! And here is the presentation of how to use CLIJ2 at one of the recent Neubias event in May 2020.

If you have an analysis pipeline built in Fiji, Icy or Matlab and processing takes a long time, CLIJ2 will help you a lot.

One-step non-toxic clearing protocol

Do you know that clearing is not just about light sheet microscopy? Even if you have done your job well and your sample is directly on the coverslip (not on the slide), as soon as your sample is thicker than 10 um (1 cell diameter), you will see the effect of the refraction index mismatch.

What is that? Your sample and the mounting medium around it have a certain refraction index (or likely several). The objective you are using is designed for a certain refraction index (e.g. air, water or oil). If these refraction indices do not match what happens? as soon as you image a tiny bit away from the coverslip, the sample will look elongated, the intensity and contrast will drop very fast.

Sounds familiar? If yes, changing your mounting medium to match the objective will solve the problem. It works for light sheet but it also works for wide field or confocal imaging! Just change your mounting medium and you will see an enormous difference!

Here is an article describing a one-step clearing protocol. This basically is about using a different mounting medium. Easy, cheap and non-toxic! Give it a try!

Have a look at this post for more info.

Multiplexed immunostaining

It is not easy to find enough antibodies that work together to be able to label a sample with more than 4 antibodies at the same time. And even 4 is pushing it.

This paper describes a new immunostaining multiplexing method called 4i. The method is based on a special imaging buffer that prevents the antibody from being strongly bound to the sample due to the imaging process. This allows the authors to detach the antibody with gentle treatments, leaving the sample in a good shape and ready for another round of labeling and imaging.

Using this method they have successfully labelled the same sample with 40 different primary/secondary of the shelf antibody pairs!

How to precisely measure the volume of a cell?

Measuring the volume of a cell is often done by labelling the cell membrane or its cytoplasm. Analysing large flat cells this way is easy but it is much harder for tiny cells like blood cells, yeast or bacteria.

Another way to measure volumes is to use a negative stain, i.e. where the medium is made fluorescent with a dye that does not go into the cell. The cell appears as a black hole in fluorescent images and unlike lipid-based membrane labelling, borders are even and easy to segment.

While many dyes can be used for live cells, one must choose large dyes when negatively imaging cells that have been fixed and permeabilized.

This paper and this one use high molecular weight (2000 KDa) Dextran to achieve these results and measure the size of bacteria.

This recent paper optimizes the technique.

 

Deep red fluorescent proteins

The microscopy field is moving away from blue dyes. This is because red light, used to excited far red and deep red fluorophores, is less damaging to live cells than near UV light which is used to excite blue fluorophores.

On top of that, red light penetrates deeper into thick samples.

So as the trend in microscopy is to move to thicker samples and use more live samples, far red and deep red fluorophores are becoming more attractive.

Here is an article describing 3 new fluorescent protein in the far red to deep red range. One can excite them with 640 nm or a 685 nm lasers or LEDs.

Matching the refraction index of live samples

To image a thick sample, it is crucial to match the refraction index of the sample with that of the immersion medium between the sample and the objective. Typically, life samples are in an aqueous solution like culture medium which has a refraction index of 1.33. Unfortunately organoids often have a higher refraction index closer to 1.44 therefore as one images deeper into the organoids, light scatters due to the refraction index mismatch and the images become blurry.

This paper presents a product that has a high RI and is compatible with cell culture. Good to keep in mind for those who image organoids over time.

How to identify cells and nuclei in an image?

NucleAlzer is a great new deep learning tool to identify roundish objects like nuclei and cells in fluorescent or bright field images.

To test if the tool works for you before you download it, you can simply upload one of your images and check the result. Easy! 😀

Follow us! :-)

Type your email address below to subscribe to our blog!

%d bloggers like this: