Good times for Illumina

Genetic analysis company Illumina has just reported a very good quarter: a 66% increase in both revenue and profits. What are they doing right?

I've posted a number of times recently about the wave of genome-wide association studies (GWAS) that has surged through the field of human genetics over the last two years. Put simply, GWAS involve looking at up to a million sites of common genetic variation throughout the genomes of thousands of disease patients (cases) and healthy people (controls), and identifying variants that are more common in cases than controls.

Although I've tended to focus somewhat on the failures of the GWAS approach (see the notable examples of height and bipolar disease, and a list of the major pitfalls of GWAS), it's worth noting that this approach has nonetheless yielded more information about the genetic basis of common diseases and complex traits in the last two years than we learnt from decades of previous research. A recent review of the field listed nearly 100 genetic variants that have been reliably associated with risk for a total of almost 40 different common diseases or traits, compared to a mere handful of well-validated genetic risk factors prior to the GWAS era - and those numbers are already seriously out of date, with new GWAS being published at an amazing rate.

The technological magic that has enabled the GWAS explosion is the development of genotyping chips: tiny glass chips covered with probes, which - when bound to fluorescently tagged DNA - provide extremely accurate read-outs of the sequence at hundreds of thousands of common variable sites (called SNPs). The picture on the left shows the Illumina Human 1M-Duo chip, which can simultaneously analyse over one million SNPs from two different samples in a single run.

The two major rival chip-makers on the market for quite some time have been Affymetrix and Illumina, both of whom produce chips to analyse genetic variation and also to look at gene expression levels (using a technique called microarray analysis).

You don't need to be a financial expert (luckily, because I'm not one!) to see that Illumina is doing something better than Affymetrix: over the last twelve months Illumina shares have more than doubled (and are currently at an all-time high), while those of its rival have halved over the same period:

(Chart from Nasdaq; both lines normalised to 100% at beginning of period.)

As I said, I'm unqualified to comment on the financial aspects of this increasingly one-sided race, but there are a few scientific issues that distinguish between the two companies. Firstly, when it comes to genome-wide SNP chips, Illumina seems to offer a superior product, at least according to an interesting paper in Nature Genetics a couple of weeks ago.

The paper, from a group at the University of Washington, set out to evaluate the performance of large-scale genotyping data (the type used in GWAS) to capture the totality of human genetic variation. They did this using full sequencing data from 76 genes in individuals from various human populations, generated as part of the ongoing SeattleSNPs project, which they could then use as a test set to determine the true coverage of each of the competing genotyping platforms.

Illumina came out of the analysis a clear winner. In the image below I've reformatted data from two separate figures from the supplementary data of the paper to make the comparison easier:

The figure compares the two current cutting-edge chips (the Illumina 1M and Affymetrix 6.0, both of which survey over one million SNPs). Each graph shows the proportion of the total genetic variation (dark lines) or common variation (light lines) that could be captured by the chips at varying levels of tagging stringency (r²) for Europeans in blue and African-Americans in red.

There's a lot of data in those graphs, but I want to use them to illustrate two points: (1) for both populations the line for the Illumina 1M chip falls more gradually than that for the Affymetrix chip, indicating that the Illumina chip provides better coverage overall; and (2) this is probably most important for the African-American samples, in whom the genetic heterogeneity of their African ancestors makes SNP-based analysis generally more challenging.

The extra coverage provided by the Illumina 1M chip was presumably a factor in the announcement a few weeks ago that the Wellcome Trust Case Control Consortium (WTCCC) - the largest GWAS consortium in the world - will be using this chip for at least part of its planned analysis of over 90,000 common disease patients, which I wrote about back in April. That sort of order must have put a smile on the faces of Illumina executives. Increased sales of genotyping chips for GWAS applications have played a substantial role in the companies' success, according to GenomeWeb News.

However, while genome-wide genotyping is currently a major feature of large-scale human genetic research, it's also a technology with a limited life expectancy: in essence, it's just a temporary place-holder for the moment when whole-genome sequencing becomes feasible. Over the next few years we will no doubt see chips with increasingly large numbers of markers crammed in, including rare variants, small insertions and deletions, and structural variation - but no matter how good these chips get they will be rendered completely obsolete by cheap sequencing technology. Why be content with assaying one or two or even ten million markers, when you can have the sequence at every position in your genome for a few thousand dollars?

Cheap sequencing also threatens the other major application of chip technology, microarray analysis of gene expression: using a new approach called RNA-Seq, researchers can now effectively use large-scale sequencing technology to read out which genes are being expressed in a tissue, and at what levels. The advantages of this technology over traditional chip-based microarray are considerable: the approach provides information on essentially every gene expressed in the cell (while microarray is limited by whether or not a probe exists on the chip for that gene); it is able to detect exactly which version (alternative splice form) is being expressed; and it has a much higher dynamic range for expression levels, which is limited for chips by the binding capacity of each probe.

A couple of weeks ago we saw an application of RNA-Seq to human cells (reported in GenomeWeb News): in an analysis of gene expression in just two cell lines, the authors identified a swathe of novel expressed regions of the genome, and over 4,000 previously unknown positions of alternative splicing (a mechanism used by genes to make multiple proteins from the same DNA sequence). It's clear from the power of this technology that the days of chip-based microarrays are numbered.

Luckily, Illumina has a backup plan. Even as its chip business has grown, Illumina has been fostering the very technology that will make its own chips redundant: next-generation sequencing. Illumina's range of next-gen sequencing machines, the Genome Analyzer series (left), represent one of the big three platforms currently competing furiously for the large-scale sequencing market. In addition, this week the company announced the acquisition of a second sequencing technology, a long-read pyrosequencing platform developed by Avantome.

Affymetrix also has a large-scale sequencing technology, using a chip-based approach - but the limited scale and inflexibility of this system means it has been rapidly overshadowed by the more powerful next-gen sequencing platforms.

Anyway, it's clear that the genetic analysis industry is currently at an inflection point, with chip-based technology still on the ascendant but visibly doomed by the simultaneous explosion of next-generation sequencing. Illumina seems well-placed to make a successful - and profitable - transition through this chaotic period.

Am I being over-optimistic about Illumina's success? Let me know what I'm missing in the comments.


Subscribe to Genetic Future.