Will the availability of open-access software for high-content image analysis change the way high-content screening vendors market their technology?
The high-content screening marketplace will have to wait until at least late fall to find out. Around that time, the Whitehead Institute for Biomedical Research plans to tentatively make available its CellProfiler open-access image-analysis software.
The software was developed by Anne Carpenter, a postdoc in David Sabatini's lab at Whitehead, soon after she joined the lab in 2003. It is designed to analyze the high-density cellular arrays for functional genomics the lab had previously developed. However, seeing a need for easy-to-use and flexible image-analysis software in high-content screening of any kind, Carpenter began to adapt the software (see CBA News, 6/8/2004).
As reported earlier this year by BioInform, a CBA News sister publication, CellProfiler was set to be released under a GPL license this spring, along with the publication of a scientific paper on the software's capabilities (see BioInform, 1/24/2005).
However, last week Carpenter told CBA News that the release of the software has been delayed until "sometime in the late fall." According to Carpenter, she and Thouis Jones, a PhD student in computer science at MIT, are still tweaking the software to make it more user-friendly.
"It is currently only appropriate for people who are knowledgeable in programming," Carpenter said. "The improvements are mostly being made in the user interface."
Because the software is not going to be a commercial product, there is no particular rush to make it available, especially prematurely. Last year Carpenter said the software would be free to non-profit users and available to for-profit entities for a nominal fee, but last week she said the software would be available for free to everyone in the scientific community.
"We have no intention of generating any profit from the project," she said. "It's intended to be open source forever. We're hoping to gain funding from the NIH and other sources to support the maintenance and development of the project itself, but there are no intentions to charge any one commercial or academic [entity] for access to the software."
Testing the Goods
Currently, various undisclosed academic and commercial entities are beta-testing CellProfiler. CBA News learned last week that one of those testers is Danish drug-discovery firm BioImage.
"Basically they're evaluating some of our cell lines, and we're evaluating their software," Len Pagliaro, BioImage's vice president of business development, told CBA News.
BioImage already had a broader collaboration in place with the Sabatini lab having to do with cell-based siRNA screening, and the two groups will be presenting data this September at the Society for Biomolecular Screening conference in Geneva.
Outside the beta-testers, other commercial entities have also been tracking the progress of CellProfiler.
"Actually a lot of companies have been interested in the project just to see how it is progressing," Carpenter said. "Some hardware companies have been interested in the software being available because they could then sell a microscope without software and point people to CellProfiler as being one option for analyzing images."
Carpenter declined to identify companies that have expressed interest in the platform, but there are several examples on the market of automated imaging platforms for cellular analysis that are not known for their image-analysis software. These include PerkinElmer's ImageTrak and UltraView platforms; and recent market entries CellWorks from Applied Precision, and NovaRay from Alpha Innotech.
Furthermore, some well-established HCS vendors have started to design cellular screening instruments that are intended to remain fairly agnostic when it comes to image-analysis software, such as Molecular Devices' ImageXpress Micro (see CBA News, 6/20/2005).
"Some hardware companies have been interested in the software being available because they could then sell a microscope without software and point
people to CellProfiler to being one option for analyzing images."
Competitive or Complementary?
Potentially benefiting from CellProfiler once it is available are big HCS players, such as Molecular Devices, Cellomics, Evotec Technologies, BD Biosciences, and Beckman Coulter; and commercial image-analysis vendors such as Definiens and Vala Sciences, Carpenter said.
"Other companies that have expressed interest are in image analysis itself, and have been interested in potentially building and selling their own modules that would work with CellProfiler," she said. "So people who were already using CellProfiler might want access to a particular algorithm that is proprietary, and then they could buy it from some company and pop it into the free software. I think this is a neat solution because it provides some incentive for people to be developing new algorithms.
"And then I think companies that sell HCS readers that come with software are curious to see the advances that are in CellProfiler, because as soon as it's open source, I presume they'll be taking advantage of some of that code," she added.
The question looms, however, as to whether CellProfiler will be seen as a competitive threat to HCS vendors. Although the big players tout differences in the quality of images their HCS platforms produce, most have come to realize that what their software does with those images is the real distinguishing feature.
"Certainly there are differences in the hardware, the most obvious being confocal versus non-confocal," Carpenter said. "But overall, the hardware for automated microscopy is a fairly mature field. So instruments have gotten to the point where they're all fairly reliable, and they're all within the same range of speed and reliability. Some are certainly faster than others, but you pay more."
Some within the industry think that HCS vendors may be forced to integrate a platform like CellProfiler into their own.
"The open source software might throw a real monkey wrench into the business of high-content screening," said one recently hired executive from a high-content screening firm who asked to remain anonymous because he is sensitive about his new position. "If you're going to win as an HCS provider, if people are going toward that open source you're going to get that open source, you're going to modify it, you're going to put it on your instrument, and you're going to make it work."
Pagliaro and Arne Huidorn, a scientist from BioImage, somewhat agree, but believe that it is much too early for commercial image-analysis software to sing its swan song.
"Historically, if you look at other image-analysis software, for all purposes, this whole process was done 10 years ago," Huidorn said. "If you take the Open Microscopy Environment, for example, you still have commercial software providers. I don't think that they necessarily exclude each other. Perhaps eventually both will be usable, and the commercial software will have some assets that the non-commercial software doesn't have."
"I think that whenever you're in the early stages of a field like this, if there's an instrument vendor that has instrumentation and sees itself as a solution provider with software, algorithms, and biology, that they would tend to see something like this as a threat," Pagliaro added. "That may be the case for a very brief window of time.
"It's not only going to be good for the end user, but also the providers, because people will get comfortable and say, 'I'm using a Ford, but it turns out you can use a Chevy fuel pump on it in a pinch, and it's not a big deal.'"
"But I think in the bigger perspective, it's going to be good for everybody in the field," he added. "It's not only going to be good for the end user, but also the providers, because people will get comfortable and say, 'I'm using a Ford, but it turns out you can use a Chevy fuel pump on it in a pinch, and it's not a big deal. I'm not going to worry so much about getting a Chevy now.' I think that's what's happening in the field."
Huidorn also made the point that CellProfiler will likely be more useful for off-line analysis of images, rather than the integrated on-line, on-the-fly image analysis that many commercial packages offer.
Furthermore, as it stands now, CellProfiler will only be an image-analysis solution, "and is really providing only one part of the entire pipeline" from image acquisition to informatics, Carpenter said.
"At the moment it is more likely to be just a replacement for that part of the pipeline if somebody already has the other areas covered," she said. "So, for example, if you're already running a commercial instrument and processing some fairly simple phenotypes using its software, you might use CellProfiler for a more challenging phenotype or cell type that you just can't get the regular software to work on."
According to Carpenter, however, some other projects are underway at Whitehead to provide, via open-access, other parts of the HCS pipeline, including integrating the Open Microscopy Environment, and developing a data visualization package called CellVisualizer. Although these initiatives are in their infancy, "maybe five years down the road there will be a full pipeline available in open source for each step of the process," she said. "Then I think it will start becoming feasible to replace some of the current offerings."
Ben Butkus ([email protected])