Skip to main content
Premium Trial:

Request an Annual Quote

Researchers Compare BGISEQ-500 to HiSeq, Microarray Platforms for Sequencing MicroRNA

Premium

This story has been updated to include comments from BGI.

NEW YORK (GenomeWeb) – In a first look at BGI's desktop sequencer, BGISEQ-500, researchers from Saarland University in Germany tested it for sequencing microRNA, comparing its performance with microarrays and sequencing with Illumina's HiSeq instrument.

The researchers published the study in Clinical Epigenetics in November. Andreas Keller, chair of clinical bioinformatics at Saarland University, told GenomeWeb that his laboratory focuses on developing non-coding RNA-based biomarkers, and is interested in developing clinical tests based on either a next-generation sequencing platform or microarray. So, he wanted to compare the platforms before deciding on which one to use in developing the tests.

For this study, BGI performed the sequencing on the BGISEQ-500, a desktop sequencer that is based on Complete Genomics' core technology that makes use of combinatorial probe-anchor synthesis and DNA nanoball arrays. The BGISEQ-500 platform is now available commercially both in China and internationally, a BGI spokesperson confirmed. Sequencing on the Illumina HiSeq was also outsourced, Keller said.

Keller said that he had two main goals for the study. First, he said he wanted to assess the reproducibility of the BGISEQ-500. "For diagnostic tests, you need stable and reproducible results," he said. "If you measure the same sample several times, you need to get the same result."

The second aim was to see whether BGISEQ-500 yielded the same results as an orthogonal technology. "It's important for us to know whether the [miRNA] signatures we discover are the same if we switch to another technique," Keller said.

In general, the researchers found that the three platforms all performed well, although each had its own advantages and biases. The group sequenced microRNA from brain and heart tissue, as well as blood.

Ali Torkamani, director of genome informatics and drug discovery at the Scripps Translational Science Institute who was not affiliated with the study, told GenomeWeb in an email that the researchers demonstrated "good reproducibility of data produced on the BGISEQ-500 platform and reasonable and expected levels of correlation between the BGISEQ-500 and other platforms."

The study "represents a first glance at an efficient and viable alternative to DNA sequencing" technology, he said. "Hopefully, we will continue to see competition and progress in this space."

In the study, BGI sequenced microRNA from six brain, two heart, and two blood samples on the BGISEQ-500, generating a median of 30.1 million reads per sample, 24.1 million of which mapped to the human genome. Of those, 23.3 million mapped to annotated miRNAs, and the remaining 700,000 reads contain potentially new miRNAs.

Looking first at technical reproducibility of the BGISEQ-500, the researchers compared technical replicates from the six brain samples, finding a median correlation of .98.

They then compared the BGISEQ-500 with an Agilent microarray finding a correlation of .48. Keller noted that this level of correlation was consistent with how other sequencing technologies compare to microarrays.

Looking closer at the data, the researchers found that while many miRNAs were measured at comparable levels on both the BGI and microarray platforms, a subset of non-coding RNAs displayed higher expression on the microarray. In particular, there were three miRNAs that differed very significantly. One was barely detected by the BGISEQ-500 platform but had 900,000 reads on the array. Another had 51.6 reads on BGI's system, but 1.9 million on the array. A third had 343.2 reads on the BGISEQ-500 versus 1.3 million on the array.  

Microarrays and NGS systems inherently have "different sensitivity levels," the authors wrote. In addition, "cross-hybridization of miRNAs with similar sequences on the microarrays or bias in library preparation" could have also played a role.  They also noted that similar discrepancies were seen when comparing the HiSeq system with the microarray.

Next, the researchers compared blood samples that they had previously sequenced with the HiSeq and analyzed via microarray to a set of different blood samples sequenced on the BGISEQ-500. Since they were looking at different samples, they focused their analysis on a subset of 2,525 miRNAs that were profiled on all three platforms as well as 658 miRNAs discovered on all three platforms.

Even though the researchers analyzed the same samples with the microarray and HiSeq, they found that the HiSeq and BGISEQ-500 data actually had the strongest correlation at .75. Correlation between the HiSeq and microarray was .6.

Keller noted that the platforms each had their own advantages and disadvantages. For example, he said, digging down into the HiSeq blood sample data, the researchers found that around 90 percent of the sequencing reads aligned to one miRNA. And in fact, 97.5 percent of the sequencing reads matched to the top four most abundant miRNAs.

When the researchers looked at the BGISEQ-500 blood data, they found a slightly different pattern. Similar to the Illumina data, four miRNAs comprised the vast majority of the sequencing reads, but at 86.9 percent of the total, the effect was not quite as pronounced. The most abundant miRNA comprised around 46 percent of the reads, while 20 percent of the reads aligned to the second most abundant miRNA, and 13 percent to the third most abundant miRNA.

Keller said that one reason for these differences may be in the amplification differences in the two technologies. The BGISEQ-500 uses linear amplification, while the HiSeq system relies on exponential amplification, which could cause more biases, preferentially amplifying the miRNAs that are the most abundant.

Nonetheless, he said, there are also some microRNAs that the group was able to detect better on the HiSeq than the BGISEQ-500. "The technology is fundamentally different," he said.

In addition, he said, the HiSeq benefits from an easier to use and faster workflow. Library prep for the BGISEQ-500, meantime, took around three days while the sequencing itself took one to two days.

Torkamani agreed that based on the publication, "library preparation time with [the] BGI technology appears to be significantly more intensive than library preparation for [the] Illumina technology."

BGI has previously said that it would incorporate automation into the library prep for the BGISEQ-500, and that turnaround time from input to results would take 24 hours. A BGI spokesperson told GenomeWeb that it has developed an automated library prep platform dubbed BGISP-100, but that it is currently not available for microRNA sequencing applications. BGI is however, running the automated library prep for its exome and whole-genome sequencing services, as well as its noninvasive prenatal test, NIFTY, and for preimplantation genetic screening protocols. 

In the Clinical Epigenetics study the researchers generated 32 gigabases of data on each of the two flowcells on the BGISEQ-500 using single-end reads of 50 bases. The researchers reported that it cost them around $200 per 20 million reads, which Keller said was comparable to their costs for outsourcing sequencing on the HiSeq.

In an ideal world, Keller said that miRNA analysis would be done by two orthogonal technologies, instead of just one. "No single technology alone is the best," he said.

Eventually, he said that his lab is looking to develop clinical tests based on miRNA signatures. The group has already validated a 12-biomarker signature for Alzheimer's disease and is also working to develop a an early lung cancer detection test based on a miRNA signature.