1. Overview Gene Expression Portfolio

As presented during the meeting, Illumina is offering chips/reagents for whole transcriptome and focused gene subsets analysis.

The whole transcriptome BeadChips we offer currently have two formats:

1) Sentrix BeadChips with 8 identical arrays (each 24K), only well annotated RefSeq Genes

2) Sentrix BeadChips with 6 identical arrays (each 48K), well annotated RefSeq Genes + less annotated genes from Unigene or EST’s

The BeadChips are available for human and mouse (rat chips will be launched in March/April).

Prices are as follows:

24K (human or mouse): $800 per chip ($100 per array), without labeling ($30 per array)

48K (human or mouse): $960 per chip ($160 per array), without labeling ($30 per array)

For the custom focused BeadChipss, the maximal # of genes per array is 1400 (for human) and 700 (for non-human). Depending on the # of samples you plan to analyze during your study we offer chips with 96 arrays (Sentrix ArrayMatrix) or with 16 arrays (Sentrix BeadChips).

Sentrix 16 BeadChip: $960 per chip ($60 per array), without labeling ($30 per array)

Sentrix 96 Array Matrix: $4,800 per chip ($50 per array), without labeling ($30 per array)

One time design fee = $12,000 -> Break even in favor for focused chips compared to whole transcrptome chips = 300 samples

Recently we have launched an assay (DASL) that is able to detect mRNA from highly degraded RNA samples (paraffine block or FFPE).

We believe having the following advantages over competitor systems:

• less starting material (50-100 ng total RNA)
• less input material for hyb (1,5 µg labeled cRNA)
• accuracy (each single probe is 30x represented on every array)
• simple and robust protocol (no fragmentation steps)
• multiplexing possibilities (reproducibility)
• most recent chip content and up to date gene annotation
• significantly less array/reagent costs

Please find enclosed some brochures regarding the gene expression products.

2. Comparison Affymetrix vs. Illumina Data Quality

Here are some background information to the comparison study:

In an effort to see if the Illumina and Affy GEX platforms produced comparable data, Barnes et al. took two samples (blood and placenta) and diluted one into the other.  All samples were then run on each platform (Affy Human U133 Plus 2.0 vs. Illumina pre-release HumanRef-8) and the results were compared.

...and explanations to "negative" statements in the publication with respect to Illumina:

1) “We first found that within-platform ‘reproducibility’ was substantially lower on the Illumina array than for Affymetrix or between-platform reproducibility (9.5% of 4312 correlations over 0.8 on Illumina; 27.2% of 30 384 comparisons for Affymetrix, compared with 24% between platforms; see Supplementary Data for details). We interpret this as indicating that the Affymetrix array contains more probe sets that are ‘truly redundant’, at least as reflected in our tissue samples.”

            Barnes et al. are defining “reproducibility” in a very strange way.  What they are essentially saying is this:

We looked at cases where each platform has multiple probes (probe sets in the case of Affy) that are designed against the same gene (but not necessarily the same transcript).  We found that the multiple Illumina probes were more likely to give different answers (when looking at differential expression between blood and placenta) than were Affy’s multiple probe sets.

            Why is this the case?  Because Illumina only has multiple probes per gene when we’re targeting alternate transcripts of the same gene.  We fully expect that alternate transcripts may have different expression patterns, so the low “reproducibility” is completely expected – it’s biological, not technical.  Affy, on the other hand, simply has a lot of completely redundant probe sets – those that target the exact same transcript and, therefore, should have the same expression pattern.  Affy’s redundant probe sets are simply a waste of array real estate.

2) “Pilot studies indicated that background subtraction had a negative impact on the Illumina data quality, so we used data that had not been background subtracted.”

            This statement stems from a common misunderstanding about background subtraction.  When subtracting background, by definition, half of the “off” genes will have a negative value, something that confuses people when they try to log transform the data (you can’t take the log of a negative number).  Additionally, it causes the data to “flare” at the low end when plotted in the log scale (e.g., in log space there is a big visual difference between 5 counts and 15 counts).  Affy has “fixed” this problem by incorrectly subtracting background, effectively adding an artificial offset. 

3) “A potential remaining source of ‘disagreement’ could be differential cross-hybridization. Similar to transcript specificity, cross-hybridization is difficult to analyze computationally because it could involve platform-specific differences that might not be reflected in the probe sequences alone (e.g. synthesis efficiency on the Affymetrix platform, or the effect of the probe identification sequences and linker for Illumina probes), and other unknowns such as the impact of highly expressed but weakly cross-hybridizing transcripts.”

            Barnes et al. speculate that our Address sequences may be causing enough cross-hybridization to affect the analytical results.  However, we have demonstrated in a peer-review publication (Kuhn et al. 2004, Genome Research 14:2347-2356) that arrays containing probes with only the address sequences pick up an average of only five counts when hybridized to labeled mammalian cRNA.

...and other notes:

1) This study was performed as a CSE and used a pre-release version of the HumanRef-8 product.  This is not the same beadset that is used in our current HumanRef-8 product.

2) Barnes et al. state that 100ng of total RNA per sample was used on the Illumina platform, but they neglect to state how much was required for the Affy platform.  In general, the Affy platform (for a single round of amplification) requires at least 1ug, usually more.

I also have attached a publication about wrong and misleading gene annotation on GeneChips (Dai et al.).

3. Illumina's Role in the MAQC Trial

A wealth of information is available to you on the official MAQC website:  http://www.fda.gov/nctr/science/centers/toxicoinformatics/maqc/

Per the website, the purpose of the MAQC project is to provide quality control tools to the microarray community in order to avoid procedural failures and to develop guidelines for microarray data analysis by providing the public with large reference datasets along with readily accessible reference RNA samples.

The MAQC group met in early December in Palo Alto to hear updates on the data analysis.  The meeting was very well attended (over 100 participants, compared to 20 for the first meeting and 40 for the second meeting) and there was general enthusiasm for the data.  The talks were all pretty superficial (no one really had enough time to do a proper analysis), but the general picture was that Illumina and Affy did better than all the others.

The current plan is to have a smaller meeting in March in Boston to dig into more data analysis.  Leming Shi, the project leader, has a tentative plan for about 10 manuscripts to be published in a special issue of Nature Biotechnology (most likely as a supplement).  The current timeline calls for submitting the manuscripts in May and publishing in the fall.  It is Leming’s intention to make the data public at the time of submission.  However, as there is such a strong interest in the data, he is working on a method for early access to the data.  Feel free to contact the project manager directly for the actual status: Leming Shi (+1-870-543-7387, Leming.Shi@fda.hhs.gov) or visit the MAQC website: http://edkb.fda.gov/MAQC/.

4. Specifications Illumina BeadStation

I have enclosed the specifications for our basis system 500X (Gene Expression System) to this email.

I have informed my colleague Mita Mancini about Dr. Cornelius Gross’s interest in the Mouse MD Linkage Panel and asked her to contact him.

As discussed during our meeting I will inform you when knowing the DKFZ decision. I would then love to start planning the scientific seminar at the EMBL Heidleberg and to talk about next steps.

Kind regards,