Prospectors get a new tool

There’s no point prospecting for gold if you don’t have a decent pan. Likewise, using your flow cytometer to collect ever larger datasets is useless if you’re lacking the tools to make sense of them. Gillian Byrne sees a new generation of analysis software as the glitter in them there hills.

FLOW cytometry - a technique for cell characterisation popularly used in clinical diagnosis, basic and translational research - was once a long-winded process, since conventional flow analysers could collect data from just 1000 - 3000 cells per second. This is no longer the case because the newest flow cytometers can create their own cellular ‘gold rush’ - sending up to 80,000 cells per second streaming past their detectors. Not only can these new instruments create data sets containing five to ten million events, but they can detect up to 18 different fluorescent labels per cell. Solid-state lasers of variable wavelength are now available, and the number of fluorescent labels that can be detected depends on how many coloured lasers are installed on the flow cytometric hardware.

These innovations have meant that, whereas researchers once struggled to force a trickle of samples through their flow cytometers, their biggest problem now is sifting interesting nuggets from a torrent of data. Unfortunately, until recently, using flow cytometry analysis software was like panning for gold without a decent pan – it wasn’t up to the job. The major problem with first generation flow cytometry software was the speed of analysis. Analysing 50 data files, each collected from 250,000 to 2.5 million cells and containing eight parameters, could take up to five full days. Visualising complex datasets was out of the question because it would take too long, and it was hard to get a clear picture of a dataset containing numerous parameters. This was because each combination of variables had to be plotted individually as it was too processor-intensive to view every possible histogram at once.

Fortunately, a new generation of software, which promises to help researchers to make valuable discoveries faster, has ridden into town. The key to its success is its ability to take advantage of the multi-core processors installed in modern desktop and laptop computers, and Microsoft’s 64-bit XP and Vista operating systems. By using parallel processing and accessing large blocks of random memory directly, software like Applied Cytometry’s VenturiOne package can work up to ten times faster than traditional systems. Faster speeds, in turn, mean that this software can include numerous new functions and use graphic-heavy interfaces that would have slowed older software to a crawl. For example, for any dataset, every possible 2-D plot can be displayed simultaneously, meaning ‘solid gold’ study results can be instantaneously spotted amid the sand.

Comparisons between VenturiOne and three older flow cytometry analysis

“These innovations have meant that, whereas researchers once struggled to force a trickle of samples through their flow cytometers, their biggest problem now is sifting interesting nuggets from a torrent of data”

software packages show the advantages this new generation of software provides. Tests were performed on two datasets containing results from 250,000 and five million cells respectively, each tagged with 11 labels. The tests looked at two main functions: compensating the data for overlap between the lasers; and creating and manipulating multiple gates. VenturiOne is the only package that automatically compensates the data – the other packages required manual compensation, in one case with the aid of a software “wizard”. VenturiOne processed the smaller dataset (250,000 events) in 1.4 seconds, while the three older software packages took between 9.0 seconds and 12.7 seconds. Two of the older software packages were unable to process the larger dataset (five million events) due to memory errors or other performance issues. The remaining package processed the data in 1.27 minutes, compared to VenturiOne’s 3.8 seconds.

Figure 1: Example screenshot from VenturiOne showing thumbnails on preview panel

Another set of tests examined how quickly the software could alter or update a workspace containing 33 plots for each dataset and, additionally, where the data had been filtered and subsetted using three levels of Boolean “AND” ‘gates’. The first test altered one of the gates and timed how long the software took to update all the plots. The second test examined how long the software took to change the gates on all 33 of the plots. Tables 1 and 2 present the results for VenturiOne and the older software.

The final test examined how easy it was to generate a workspace for basic analysis. This workspace contained 37 plots gated on A with quadrants. The time taken to generate the workspace and the number of mouse clicks required to perform the functions necessary was examined, as shown in Table 3. Again, the new software outperformed the old by a considerable margin.

Having access to fast, simple software, such as VenturiOne, has made a significant difference to researchers. Albert and Vera Donnenberg, researchers at the University of Pittsburgh, describe how analysing 50 data files of 250,000 to 2.5 million cancer stem cells and eight colours using an old software package took up every morning of a ten-day holiday. They were often frustrated by software crashes and excruciating refresh intervals1.

Along with faster number crunching, the reduction in the number of clicks required to perform any function means that it is easier to train less experienced staff and students to use the new software. Ludovic Zimmerlin, a second year PhD student in the Donnenberg’s lab, describes how access to the Donnenberg’s VenturiOne software has helped him to separate very rare adipose progenitor cells from noise because he can preview his results simultaneously. Since any of the twelve investigators working in the laboratory could be competing for time on the software, during a two-day period, it is crucial that it allows him to make discoveries quickly. Teaching newer students, is also simple for Ludovic since he only had to watch the Donnenbergs using the software twice before he could operate it. The software is designed to be intuitive, especially the preview panel, which allows every combination of two parameters for up to 14-parameter datasets to be viewed simultaneously (see Figure 1). Interesting data can easily be picked out from the panel, just like searching picture thumbnails on your computer for the right photograph.

As flow cytometers have grown faster and increased the number of colours they can use, researchers have found it difficult to sift out the really valuable results. New software is helping them strike gold in their research by working faster and being easier to use. Prospecting at the frontier of research will never be the same again.