The co-founder and former chief technology officer of now-defunct high-content screening start-up Vitra Bioscience has established an online repository of normative images for important cell-based assays that he hopes will further the development of standards for the HCS industry, CBA News learned this week.
The web portal now seeks contributions from the HCS industry in the form of donated cell-based assay images — which have thus far been difficult to come by due to a lack of participation from HCS vendors and the technical challenges associated with running the same assays on a variety of instruments.
The former official, Ilya Ravkin, established the image library a little over a year ago under the auspices of the Data and Image Analysis special interest group of the Society for Biomolecular Sciences, which he co-chairs, Ravkin told CBA News this week.
Ravkin said the library's goals are to validate and compare current image-analysis algorithms; compare imaging instruments and modalities such as confocal, widefield, and laser scanning cytometry; help researchers implement HCS assays by presenting typical images using different assays, cell types, and compound concentrations; and facilitate the adoption of a common standard for exchanging high-content analysis images and metadata.
"All major instrument manufacturers ignored or explicitly refused to take part in the algorithm comparison."
When he started the portal, Ravkin had the backing of Vitra, and at SBS' annual conference in September 2005 his company-sponsored poster entitled "Comparison of Several Classes of Algorithms for Cytoplasm to Nucleus Translocation" won the best poster award in the Imaging Technologies category.
Since that time, however, Vitra ran out of cash and went out of business (see CBA News, 11/28/2005) and said it would sell its flagship CellCard technology to Serologicals (see CBA News, 2/24/2006). The status of that transaction remains unclear following Millipore's April announcement that it would acquire Serologicals.
Ravkin now hosts the SBS special interest group and high-content image library on his personal web site, which can be seen here. Ravkin will again co-chair the Data and Image Analysis special interest group at SBS' next conference, to be held Sept. 17-21 in Seattle, and hopes to make more of his peers aware of his library.
Of the goals Ravkin outlined for the library, he said that facilitating the adoption of a common standard for exchanging high-content analysis images and metadata "is probably the most difficult to achieve, but also the most urgently needed in this area. High-content analysis is suffering from a lack of interoperability, open standards, and numerical characterization of the quality of different commercial offerings."
This goal has been difficult to achieve in part because it's tough to appropriate image sets from unwilling HCS vendors, but also because of the technical difficulties of scanning or imaging the same well plates on different instruments, Ravkin said.
"The past year of the image library has been a mixed success," Ravkin said. "On one hand there is much interest in algorithm comparison … and I am continuously getting requests for images from different institutions.
"On the other hand, all major instrument manufacturers ignored or explicitly refused to take part in the algorithm comparison," he added. "It was hard and took a lot of time to collect the images that we have now, and … it is equally hard to expand the library."
Ravkin said he invited vendors to compare algorithms during last year's SBS special interest group. Although "all major vendors were on the [invitation] list" according to Ravkin, he conceded that he wasn't sure he targeted the appropriate people at those companies.
In particular, Ravkin said that he had contacted a manager at one well-known HCS vendor who told him that he "liked the idea in principle, but was advised not to participate at that time." Ravkin declined to disclose the identity of this vendor. He also noted that CompuCyte thus far has been a contributor and has shared images and results from a Tranfluor assay conducted with collaborator Olympus.
"High-content analysis is suffering from a lack of interoperability, open standards, and numerical characterization of the quality of different commercial offerings."
CBA News this week called or e-mailed several well-known HCS vendors regarding Ravkin's claims of non-participation. GE Healthcare and BD Biosciences did not respond in time for this publication; Molecular Devices declined to comment.
Judy Masucci, Cellomics' director of marketing, and Mark Collins, Cellomics' senior product manager for bioinformatics, told CBA News that to their knowledge Cellomics had not been contacted by Ravkin regarding the image and image-analysis comparisons. They added that several scientists at the company — in particular, Collins — were aware of Ravkin's work and that Cellomics has been an ongoing supporter of the creation of HCS standards.
"By promoting standardization for data interchange and so on, you actually make the market bigger," Collins said. "An open standard makes a huge degree of sense from our perspective, and that's why we're behind it. The quicker you can get a standard together, the quicker people can build products based on that standard and all make money."
Cellomics and Ravkin both also mentioned the participation of the National Institute of Standards and Technology. John Elliott, a biotechnology research scientist at NIST and member of a recently formed NIST sub-group that specifically deals with cell and tissue measurements, generated a lot of interest with a standards presentation at a Cambridge Healthech Institute HCS meeting earlier this year. Ravkin and Collins both said that there was discussion of a NIST-brokered standardization effort, but since that time, however, they have not been in touch with Elliott.
In a February interview with CBA News, Elliott recapitulated NIST's desire to facilitate the development of standards in HCS and automated microscopy, in general. At that time, he said there was no timeline for the development of a standards consortium, and that its development was very dependent on vendor participation. Calls made this week to NIST were not returned in time for this publication.
Much of the early participation seems to be coming from academic high-content screening scientists. For instance, Ravkin said that the CellProfiler open-access image-analysis group at the Whitehead Institute is planning to publish results of analyses of the current images sets using the CellProfiler software.
In addition, independent image-analysis software vendors may have more of a stake in developing standards than platform vendors, and as such have also expressed interest. In fact, Kurt Scudder, an application field scientist for software firm Definiens, is chairing the SBS special interest group along with Ravkin.
The online library currently contains two sets of images: cytoplasm-to-nucleus translocation assays obtained using Vitra's CellCard reader and GE Healthcare's IN Cell Analyzer 3000; and Transfluor assays obtained using a CompuCyte iCyte imaging cytometer and a Cellomics ArrayScan reader.
Currently, only the assay images are housed on the site. Accompanying metadata — for example, drug concentrations in different wells — is at the moment in a variety of forms such as Word documents, Excel spreadsheets, and plain text documents, according to the website.
"At the moment, metadata is provided with images in ad hoc formats," Ravkin said. "My next project will be to provide the library with metadata converted to a format that has a chance of being adopted as a standard."
Though far from being complete, the "mere existence of a publicly available image library for important classes of assays has a formative influence on the field," Ravkin added. "A limited and imperfect library is a lot better than no library."
The SBS Data and Image Analysis special interest group is currently soliciting topics and panelists for September's meeting, and the web portal continues to solicit donations to the image library.
"Acceptance of this invitation indicates agreement to make your results public," the website states. "Originally we were planning that the [special interest group] chairman would collect the results, calculate statistical quality metrics, and arrange results for presentation [on] the SBS website and/or for publication in a magazine.
"While we still hope to be able to do this in the future, a more practical approach has emerged," the site continues. "Participants in the algorithm comparison publish their results independently referencing this image library. We will provide on this page links to such publications and presentations."
— Ben Butkus ([email protected])