Right, and basically that's what we're using. Except, there are issues with doing this for individuals instead of businesses or setup labs. First, labs will not provide any method for collection. It's typically up to you to go get blood drawn or attempt to get a sales rep at a supplier to sell you a collection kit. Also, it's then up for you to do international shipping of the samples with customs clearance, etc. And, finally, most of these labs then require you to mail a hard drive along with the sample and then they mail that back to you for an extra fee.
It's also unlikely that most of these labs will offer to individuals at all. Why? Huge compliance and regulatory risk. Being a lab that processes samples is one thing, offering a 'consumer' or 'hobbyist' kit is something else entirely and brings with it a whole boat load of lawyering required (something we're already doing).
If you work in a lab now and are fine negotiating contracts with different suppliers and are used to mailing biological samples, etc... then absolutely this is doable for you.
Also, my search did not bring up that price on genohub. The lowest I saw for a single 30x whole human genome was $3,050.
We're basically doing all that leg work for the difference in price and are buying the kits and handling all the paperwork and bs.
Your approach of storing only the variants is pretty clever. But what's your contingency plan for when the reference you're using becomes obsoleted? There are plenty of options--I'm just curious to see your thinking.
The initial version will probably just be layers of variants on the reference, but I'm hoping to quickly replace that as I think it leads to unnecessary bias in the way we interpret the data (oh, this is *not normal*).
My plan is to implement a full graph approach that has no real reference, and each genome is just a path through a the graph of the whole of our stored genomes.
I like it! Thanks for the explanation.
Are you really planning to ONLY store variants? I don't think is a good idea. While people have had the dream of a graph based representation like you describe for years I have yet to see a convincing and practical solution for it (despite many smart people working on it). It is harder than it seems. But, even more importantly, by not keeping the raw reads you lose the ability to take advantage of new alignment and variant calling methods. The cost of storing a 30X whole genome worth of raw unaligned reads (~80GB compressed) is relatively negligible.
Of course we will keep the bem files in "cold storage" as static files. But, the system to live query them has to be far more efficient and that's what will be variant based.
I think you probably should have done some trial runs and cooked up some demos using your own sequence before you launched this Kickstarter.
Actually, that was part of the issue... we couldn't find a lab that would work with 2 samples! They needed us to do the legal legwork, etc. Plus, sample collection was an issue. We're hoping to get the "group discount" and pay into the bulk purchase ourselves.