Edinburgh Genomics has installed Illumina’s v4 upgrade to the HiSeq 2500, the first facility to do so in the UK. Our initial trials show that the upgrade delivers to Illumina’s promises, and will indeed generate 1 Tb of raw data per dual-flowcell run. This will rapidly translate into cheaper sequencing for Edinburgh Genomics collaborators and customers.

The context

Across next generation genomics, the drive is for better data, more data and reduced costs. Illumina recently launched its v4 reagent kits for the HiSeq platform. The company promises up to 1 Tb of data per dual flowcell run of the HiSeq 2500, using 125 base reads, in 6 days. The new kits are compatible with all factory-built HiSeq 1500/2500 and HiSeq 2000 models shipped in 2013 via a paid-for upgrade.

At Edinburgh Genomics we have four v4-compatible HiSeq 2500s. So when Illumina announced the availability of the new kits, we made sure to test the capability of the upgrade as soon as it launched.

The experience

The upgrade itself was pleasantly straightforward. It took our field service engineer less than half a day to complete. There are no hardware upgrades, only software and firmware changes. This is good news as hardware upgrades are more complicated, and have been troublesome in the past.

Rebooting the instrument after the upgrade and navigating to the high output mode is a familiar experience. The only obvious change is the “HiSeq v4” banner at the top of the HCS home screen.

Setting up the run and loading the reagents is as simple as setting up a rapid run. It is no longer necessary to hold back the paired-end reagents. The instrument can now go through the turnaround steps without operator intervention. This is a welcome convergence with rapid runs, and minimises down time during paired-end, high output runs.

The new v4 high output welcome screen

Data volume

The v4 upgrade has so far delivered close to Illumina’s specifications. Our first v4 run achieved 442 Gb of raw data (467 Gb including indexing reads) from a single flowcell in under 4.5 days.

Flowcell-level run summary. Note the low error rate, and high overall bases over Q30 in the insert reads (91.0% for read 1 and 88.0% for read 4).

This is 1.5 days faster than the times advertised by Illumina for dual flowcell runs, presumably because of reduced data processing time when running a single flowcell. Illumina’s recommended loading concentrations for v4 flowcells remain unclear, but we anticipate achieving ~250 M clusters per lane at the optimal cluster density, or 2 billion read pairs per flowcell. In context, this is equivalent to ~10 human genomes at 30x per dual flowcell run, or 1.7 human genomes per day.

Data quality

Data quality was equally impressive with > 88% of bases at Q30 or more across the flowcell. At optimal cluster density, we obtained over 92% bases at Q30 or more.

Lane-level quality plot. The graph shows the percentage of base calls exceeding phred-like Q30 scores at each cycle for one of the lanes with optimal cluster density (~ 945K clusters/mm2).

In our experience, this is noticeably higher than the average v3 quality, but on par with the HiSeq’s performance in rapid mode.
The v4 software also comes with a new version of Illumina’s real time analysis algorithm that removes the need for a control lane. This is potentially a major breakthrough as sequencing of unbalanced libraries has been a long-standing issue on the HiSeq. At first glance it looks like Illumina may have finally cracked this problem (more on this in a future post).

Why we are upgrading

The upgrade adds considerable daily throughput to the HiSeq 2500. A v4 high output, 125 base paired-end, dual flowcell run can yield up to 166 Gb per day. This is 50% higher than we can achieve in rapid mode using 150 base paired-end reads, and a massive 200% higher than the daily throughput of a 100 base-paired end v3 run. Producing 1 Tb of sequencing data from a single run in 6 days is a technically impressive feat but, unlike the rapid mode, the key selling point of the v4 upgrade is cost, not throughput.

Our initial calculations suggest that switching to v4 will allow us to slash the sequencing component of our standard charges by approximately a third. This is welcome news, as the cost of Illumina data hasn’t dropped since the v3 chemistry was first introduced in 2011 - if anything prices have gone up with inflation.

Posted by Karim Gharbi with contributions from Mick Watson and Mark Blaxter