We just came back from the Great Lakes International Imaging and Flow Cytometry Association held in Troy, Michigan. While it remains a hidden gem for the greater research community, it is an inescapable meeting for core facility people in the field of flow cytometry and imaging. It reliably offers a stellar scientific program as well as the opportunity to learn about the very latest developments in the field from others attendees and vendors alike. The CAT Facility was well represented this year, with a presentation from Laura Johnston on our Spectral Flow training program during the Cytek pre-GLIIFCA User Group meeting, a poster from Bert on the identification of high-performance brewer yeast using the Image Stream and my presentation on marketing strategies to promote new technologies in a core facility. If you want a look at any of these presentations, let me know!
Once again, GLIIFCA proved to be a thought-provoking meeting. Here are three take-away ideas
- Efforts need to be put in training in computational analysis tools
With the ongoing interest for examining more and more parameters on on a single cell, researchers are turning to unsupervised methods for high-parameter data analysis. We heard from Dr. Anna Belkina, who modified the t-SNE algorithm to be able to examine more data points (https://www.biorxiv.org/content/10.1101/451690v3). With the original t-SNE algorithm, many users needed to downsample their data in order to get results. The problem with this is that important information could be lost by downsampling, especially if there are rare populations that are of interest. If large numbers of cells were run on the original t-SNE algorithm, the t-SNE plot would look like one population of cells without distinct islands. Dr. Belkina’s algorithm, opt-SNE, can handle large numbers of cells and produce a plot with distinct islands. If you would like to try out this opt-SNE algorithm, it is now the default t-SNE algorithm in FlowJo 10.6.
We also heard from Dr. Jonathan Irish, who described his clustering algorithm, Risk Assessment Population Identification (RAPID) (https://www.biorxiv.org/content/10.1101/632208v2). Dr. Irish compared his clustering algorithm to Citrus, which can be used to connect cell clusters to clinical outcomes. However, the RAPID algorithm improves on this with a more unsupervised approach. Citrus requires an expert to assign patients into groups and beforehand, whereas RAPID utilizes an unsupervised method to group patients at the end of the analysis.
Both of these talks highlighted the importance of computational analysis tools and how researchers have been working to improve these algorithms to be better suited to flow and mass cytometry data.
- More lasers, more troubles
Dr. Bill Telford presented his observations with a wave of new lasers coming out in some not-so-distant future. It seems that the far-red lasers can’t excite fluorophores all that well, so manufacturers are currently looking at Deep-UV lasers (DUV) – anything below the current UV laser (355nm). There are already some DUV fluorophores available, both single molecules and tandem dyes, all of which provides a larger group of toys to play with. So, when are we getting these upgrades? It might take a while actually. There are a few kinks that need to be ironed out first. The main one is that these lasers have an extremely short life span, say a few months or so. There are other weird snags in using these DUV lines. For one, the autofluorescence of the cells gets to ridonculously high at these low wavelengths, making resolution of a positive marker pretty hard. Second, and this is from an observation from another attendee during Dr. Telford’s Q&A, it seems that these DUV lasers might have the ability to trigger chemical reactions with certain molecules, which could turn into generating polymer noodles right in the instruments flow cell. I don’t know enough about this specific matter, please feel free to leave a comment below if you think it’s a likely issue.
- Victims and perpetrators
From Ryan Duggan, former CAT Facility Technical Director but now Senior Scientist at AbbVie, we had an eye-opening talk on the use of high dimensional spectral flow to dissect the tumor micro-environment. While discussing panel design on the Aurora, Ryan offered what I thought was a particularly useful nomenclature to describe the behavior of the fluorophores in a panel: the victims and the perpetrators. Taken from the world of chemistry, these two concepts aim at clarifying the role of each fluorophores: the victims are those who receive spillover from other markers. On the other end, the perpetrators bleed in other channels. It’s an intuitive concept that facilitate the comprehension of the interactions of the fluorophores in a panel. We’ve created a list of victims and perpetrators for the aurora on our resources page, which you can check out here.
- Meet BenchSci
Lastly, we were introduced to BenchSci. I like them a lot so far. The amount of time that I, you, and everyone we know spend on figuring out which antibody actually works is already pretty high, but the answer also changes depending on how the sample is prepared and which technology is being used. Enter BenchSci. This Toronto based group propose a web-based tool to give you all the information you can possibly want from any given antibody. Simply type the name, clone application, catalog number or whatever you have at hand to identify the antibody and the site will populate a result page with a list of figures coming from scientific papers, antibody suppliers, or independent sources. Sure, the results are always positive, you probably won’t get a notification telling you that a specific antibody won’t work. But the footprint left by the antibody on the internets will help gage how trustworthy it is. If you only get back results from the ‘Camdenton Regional Gazette of Irreproducible Results‘, it should mean something.
You can get a BenchSci account for free (you academic person you). Here’s a bit of advertisement. Have a look and let us know what you think below!