SAN FRANCISCO (GenomeWeb) – Different droplet-based single-cell RNA sequencing methods have distinct advantages and drawbacks, making each best suited for different applications, according to researchers from Tsinghua University in Beijing.
In a study published this month in Molecular Cell, the team tested single-cell RNA sequencing using 10x Genomics' Chromium system, the Drop-seq protocol, and the inDrop method, which has now been commercialized by Harvard University spinout 1CellBio.
Jianbin Wang, senior author of the study and assistant professor at Tsinghua University, said his lab focuses on technology development and wanted to compare different platforms, both for their own research and to provide a useful guide to others. To provide more of an apples-to-apples comparison, the team included the three droplet-based platforms that were available at the time — 10x, Drop-seq, and inDrop — and not other single-cell sequencing methods, like those that make use of microwells, for instance. Wang noted that BioRad has since launched a droplet-based system, ddSeq, which was not available at the time his group began its study.
To test the methods, the group analyzed a lymphoblastoid cell line, generating two to three replicates for each method. The team also developed its own pipeline for data analysis that would work for all of the systems.
"Very generally, every system is able to profile the transcriptome," Wang said, although no single platform can serve every need perfectly. In general, the 10x Chromium was more robust than either Drop-seq or inDrop and had the least amount of technical noise and variability, Wang said. That finding wasn't too surprising, given that the system has been commercialized for longer than the inDrop system and given that the Drop-seq protocol is freely available for anyone and so has not undergone extensive quality control. The Drop-seq online protocol has been downloaded more than 60,000 times with very detailed instructions and its advantage is that "it's cheap and easy to assemble," he said.
One finding that was somewhat surprising, Wang noted, was the impact of different beads. All three systems use cell barcodes attached to beads. In order to ensure that one cell is encapsulated with one barcode, the beads and cells are diluted. But the beads themselves can have an impact on how much dilution is needed, and therefore on cell capture efficiency. For instance, both 10x and inDrop use hydrogel beads, which are large and physically flexible, whereas the beads used in the Drop-seq protocol are smaller and more rigid, which increases the chance for doublets, so more dilution is required, reducing cell capture efficiency.
With the larger and more flexible hydrogel material, the speed at which the hydrogel beads can flow through the system is naturally reduced, Wang explained, so less dilution is needed, resulting in better efficiency.
In addition, the researchers found that the gene expression data tended to cluster based on the system that was used, indicating "system-specific quantification bias at the gene level," the authors wrote. Looking closer at the sources of bias in each system, the researchers found that the Chromium platform favored shorter genes and genes with higher GC content, while Drop-seq was better at detecting genes with lower GC content.
All of the methods were consistent within the technical replicates, indicating that while data from multiple experiments on the same system can be combined, comparing data generated by different techniques would be more challenging. This finding was similar to a previous study by the Association of Biomolecular Resource Facilities' Genomics Research Group, which compared platforms from Fluidigm, WaferGen, 10x Genomics, and Illumina/Bio-Rad.
In terms of performance, the researchers found that the 10x system had the highest sensitivity, capturing more than 17,000 transcripts from around 3,000 genes on average. Drop-seq detected 8,000 transcripts from 2,500 genes, while inDrop detected 2,700 transcripts from 1,250 genes. The 10x system also had the least amount of noise.
Wang said that one cause of noise in the inDrop system was variability in the beads, adding that results from two batches of beads for the same sample were "quite different."
However, the higher sensitivity and robustness of the 10x system comes at a cost. The instrument is the most expensive to purchase, with a list price of $125,000, or $75,000 for a version that runs only single-cell experiments. The researchers estimated per-cell costs to be $.50.
The Drop-seq protocol is the least expensive option, Wang said. The researchers reported that it cost them less than $30,000 to build the Drop-seq system and per-cell costs were $.10. However, Drop-seq costs for Wang's group were significantly higher than the estimated costs of $6,000 for building the instrument. Wang said that scientific equipment tends to be more expensive in China, which could explain the difference.
The main drawback for Drop-seq is that it has a low cell capture efficiency, so it is not as well suited for precious samples or samples without many cells, Wang said.
For inDrop, the group estimated the instrument cost was around $50,000 and per-cell costs were $.25. Wang said inDrop's main advantage is that it is an open system, so users can tweak and customize it for their own purposes. He noted that it is not yet as robust as 10x Genomics' instrument, but he anticipates it will improve over time.
For his own lab, which is focused on technology development, having an open system is an advantage. "We can play with it," he said. For instance, the lab is interested in designing the oligos on the inDrop system to also include antibodies, which would enable proteins to be analyzed along with the RNA. In addition, the group is considering doing its own custom analyses of RNA splicing, or performing single-cell DNA sequencing. "Everything can be tuned on this system," he said.
Jan Philipp Junker, a developmental biologist at the Max Delbruck Center for Molecular Medicine in Germany, who is an early-access user of 1CellBio's inDrop system, said that although his lab has not done an in-depth comparison of the data quality between the three systems, he agrees with the overall assessment of the pros and cons of each system. Like Wong, he appreciated the openness of the inDrop system but said that compared to the Chromium, it was "less polished when it comes to usability and barcode design." Junker's lab is now focused on using both the inDrop and Chromium systems and decided not to use Drop-seq due to its lower cell recovery.