NEW YORK (GenomeWeb) – Earlier this month, Fluidigm notified customers that the chips for capturing medium-sized cells for its C1 single-cell system were often capturing more than one cell, GenomeWeb has learned.
In addition, according to some experts, the issue may also rear its head in other single-cell analysis technologies. Although they had not yet tested other technologies, it will be something researchers should be mindful of as the single-cell market continues to grow.
In a recent entry on his Core Genomics blog, James Hadfield wrote that Fluidigm contacted customers about the problem and said that for its chips that capture medium-sized cells, 10 to 17 micrometers, approximately 30 percent of the chambers contained multiple cells.
In an email to GenomeWeb, Fluidigm spokesperson Howard High confirmed that the medium-cell chip appeared to have a "higher-than-acceptable rate of doublets" in any given chamber, but added that the company is working on a fix to the problem, and it expects to make redesigned IFCs available to customers in the spring.
The firm is "confident, based on our investigations, that these design changes will resolve the issue," High added. In the meantime, Fluidigm has published a white paper on the issue in which it ran numerous experiments to evaluate the rate of doublets. In its C1 96 IFC, it found doublets in approximately 30 percent of the sites with a standard deviation of 10 percent, while on its C1 HT IFC the average number of doublets was 44 percent of occupied sites with a standard deviation of 23 percent.
Fluidigm noted that the majority of doublets come in the form of one cell stacked on top of another within the capture site. "Although we had previously observed occasional instances with two cells in different places in the capture site, stacked doublets were considerably more difficult to detect," the firm wrote in the white paper.
The company also laid out a number of steps that users could take to identify doublets until its new chips are available. Fluidigm wrote that users should routinely image captured cells and could use either bright field/phase contrast as well as fluorescence imaging to identify doublets.
High added that the company has also been working with its customers "so they can implement recommended approaches to detect and remove data from chambers with two or more cells."
At the end of the third quarter 2015, Fluidigm had approximately 375 C1 systems installed. Fluidigm declined to disclose the impact of the issue on its future business outlook until it reports its 2015 Q4 earnings in February. The C1 system itself is "functioning properly," High added, and the design work the company is doing on the medium-cell IFCs should "address the issue for the long-term."
Hadfield told GenomeWeb that despite the issue, his lab still intends to use the C1 system, but with "more careful experiments, [with] additional verification and lots of validation." In addition, the chips that capture smaller cells "appear to be less affected so we're still planning as normal with those," he said.
Stephan Lorenz, senior scientific manager at the Wellcome Trust Sanger Institute's Single-Cell Genomics Core Facility, told GenomeWeb that most of its users work with the small-cell chip anyways, so the doublet issue in the medium-cell chip has not been a huge problem.
The reason the medium chip has a higher doublet rate is because there is "a little extra space" around the capture site, where "cells might stack," Lorenz explained. Researchers who have done experiments with the medium-cell chip will now have to go back to the imaging data to confirm that they have captured single cells.
Researchers may also "have doubts about their historic data and might be annoyed," but, Fluidigm "has dealt with it well," Lorenz added. The company "reached out to the user base to try to mitigate the problem."
Nonetheless, the issue could potentially impact Fluidigm in the future, particularly as a number of other companies enter the single-cell space. Illumina and Bio-Rad recently announced a partnership to develop a single-cell sequencing system; WaferGen recently launched its ICell8 single-cell system; and 10X Genomics recently launched a kit for single-cell RNA sequencing.
However, Lorenz said he didn't think Fluidigm would lose too many customers because of the problem. Fluidigm had the first single-cell device on the market, and it has been in the hands of customers since 2012, while other companies' products have not yet been broadly tested. In addition, "there will be more interesting products from [Fluidigm] in the future," Lorenz predicted.
Hadfield added that he did not think that Fluidigm would be the only company to have issues with capturing more than one cell. "I suspect this is going to be a problem for all single-cell systems, and the community needs to consider this and the tools we need to spot it early on," he said.
Tomislav Ilicic, a bioinformatician in Sarah Teichmann's research group at the European Bioinformatics Institute, agreed that issues with single-cell capture are likely to plague not just Fluidigm, but other technologies, as well. "I think it's a problem all single-cell technologies will have," he told GenomeWeb.
Ilicic said that he has been working on designing an algorithm that can identify problems in single-cell sequencing data. He started working on the algorithm after noticing issues with the Fludigim C1 system. He said there are three main types of problems that can occur when capturing single cells. The first is that the nothing gets captured. This can cause problems because sometimes there is contamination in the well that will generate data or leakage of material from an adjoining cell. "This doesn't happen often, but you still want to get rid of it when it does," he said.
The second issue is that sometimes the cells break. Often broken cells result in an excess of mitochondrial RNA or DNA, he said.
Finally, the third problem is capturing multiple cells. "These are particularly hard to distinguish from a normal cell without looking under the microscope," Ilicic said.
For RNA-seq experiments, one thing that researchers can do is compare gene expression of the individual wells with mean gene expression. Wells with multiple cells will be closer to the mean expression, Ilicic said. However, this comparison works best when studying a group of cells that are mostly homogenous. The more heterogeneous the cell population, the less effective this comparison will be.
Ilicic said he has identified a total of 20 such features to distinguish sites with single, healthy cells from sites with multiple cells, a broken cell, or no cell, which he has used to build a software tool. He expects that a paper describing it will be published in a peer-reviewed journal in a few weeks.
Researchers will then be able to use the software tool to parse their good single-cell data from data that is coming from multiple cells. This should help save time over manually imaging each chip, Ilicic said. Imaging chips is not a problem when examining just one or two, he said, but most researchers sequence between 10 and 30 chips, and examining each under a microscope is time consuming. With the software tool, he said that researchers should still image one chip, and use that data as the training set to calibrate the algorithm and then run the software on the remaining chips.
One unknown factor is the impact of this issue on experiments that have already been done and data already published. "This is a very hard question to answer," Ilicic said, and is largely dependent on the question the researcher is asking.
"The more multiples you have, the more heterogeneity you'll mask," he said. For experiments that attempt to analyze new genes, or differential expression of a gene between cells, it may have a significant impact on the conclusion if one of those cells is actually two cells, he said. But, if the goal is to get a more general picture of gene expression, then the impact will not be as great, he added.