Microarray Replicate vs Sequencing cost break even point.

[rskr]

As has been mentioned in several threads, Microarray slides are cheaper but require more technical replicates than RNA-seq. I wonder what the cost point is where the one or more RNA-seq runs are cheaper than the multiple Microarray technical replicates.

[mbblack]

I do not know many people who bother with technical replicates for either experiments. At least as far as gene expression studies, for arrays, the standard assumption is that technical variation is trivially small relative to biological replication, and the same is true for sequencing. Earlier experimental work showed that assumption to be valid so nobody I know usually concerns themselves with technical replicates unless the specifics of an experiment warrant it (samples prepared at greatly different times, or by different labs or some other aspect that would cause one to worry about technical variability specifically).

At least for the sort of experiments I work with, we have not bothered with technical replication in years, but we always include 4 or 5 (sometimes more, if budget allows) biological replicates.

Seems to me the main differences in cost are related to required or desired sequencing depth, and the personal time involved. Microarrays, at least Affy titan arrays, take less time than sequencing for the same set of samples, and the human labor cost is a very non-trivial part of the cost equation. Also, to get robust sequence results, the sequencing depth required may add substantially to the cost as well, depending on the scope of the experiment.

I am also continually surprised by posts in these forums of people trying to do comparative expression studies with RAN-seq who have somehow arrived at the conclusion that sequencing magically negates the need for biological replicates. Regardless of cost, an experiment has to adequately sample for biological variation, since that is the single largest source, by far, of variation in any expression study and regardless of how expression is measured.

[jwfoley]

Are microarrays really cheaper? My service center will prepare an RNA-seq library for CAD $135, and 10 million reads (assuming a mammalian genome, but low depth since you're trying to save money) on a HiSeq at 1x50 will cost about $40. However much cheaper microarrays are per sample than $175 (including the microarray technical replicates, which are not done for RNA-seq because there's no need), RNA-seq data are so much better that you cannot justify starting a new project with microarrays today.

Regarding biological replicates, see doi:10.1093/bioinformatics/btt688.

[HESmith]

I agree with mbblack that technical replicates are unnecessary while biological replicates are essential. However, I can't see how microarrays can be cheaper than sequencing.

The citation by jwfoley indicates that multiple (4+) replicates @ 10M reads each is a sweet spot for sensitivity vs. expense (at least for differential gene expression). Our reagent costs for automated RNA-Seq library prep are $90/sample, plus $55/sample for 24 multiplexed libraries (good for >10M reads each) in one single-end 50bp Rapid run, for a total of $145 each. The labor to set up the library prep and sequencing instruments is minimal.

In contrast, our cheapest per-array price is $175 for the 2.0 ST Array - that's w/o any reagents, and requiring considerable hands-on time. The Titan arrays do minimize labor, but are significantly more expensive. I'm happy to be corrected but, unless you include other factors in the calculation (e.g., informatics support), microarrays are not cost-competitive.

[DonDolowy]

We pay about 125€ per sample 2x50bp PE @ 15 mio reads. Does not get much cheaper than that!

[lchpoon]

Do those prices include library prep?

I was recently at the Illumina NextSeq seminar and surprise, surprise they pointed out how RNA seq is cheap than most arrays. I thought this was more to do with the amount of information you can obtain through RNA seq vs genotyping, 1 probe at a time; I had no idea that this is even true as an absolute cost.

[DonDolowy]

In our case it does include library prep. We just hand over the RNA.

[jwfoley]

Yes, my center's CAD $135 is the cost of library prep. You can probably get the reagents cheaper if you do it yourself, but you need a sonicator. I've been looking at a protocol that might take the library prep cost down to about $10, but it'll be a while before I know how well it works. However, sequencing costs are coming down all the time.

I was recently at the Illumina NextSeq seminar and surprise, surprise they pointed out how RNA seq is cheap than most arrays. I thought this was more to do with the amount of information you can obtain through RNA seq vs genotyping, 1 probe at a time; I had no idea that this is even true as an absolute cost.

I think this has been true for several years.