Download the whole genome of Cannabis Sativa

I don't have much interest in this dataset, but I can describe what would generally be done.

The data here comes from tiny fragments of the genome. On each end of the fragment, 100 base pairs of DNA are sequenced, and that's what are in the text files, along with confidence estimates of each called base pair. These bits of sequence are called "reads."

The first step would be to take all these fragments and assemble them into a full genome. The page notes that they have 327x coverage, meaning that on average, there will be 327 reads that overlap a given region of the genome. To assemble the genome, you find these parts that overlap, and try to expand out from there.

Genome assembly is complicated by many things, but among them are some regions that get very few fragments, and repetitive regions. If you have a 150bp segment repeated many times, often you won't know how large it is, because your reads are only 100bp, and unless you have really long fragments or varying fragment size, you won't be able to estimate the length of the repeat. Also, if the repeat occurs in other, or likely many places, you have additional difficulties matching what's on the left and right side of all these repeats. This is an active area of research, and people are building new assemblers and new assembler technologies even today, to deal with the changes in sequencing technology. It's unlikely that from a dataset like this that you'd be able to get much of an assembly, one would likely have to perform many more targeted sequencing projects to fill in the gaps.

However, with the beginnings of an assembly, the first thing to do is compare it to the genomes of related species. This will identify both the genes, and what has changed the most in the genes. One could attempt to identify which genes are responsible for interesting metabolism or other features of the species. If you have a trait of interest, and sequence other variants of the species, you can perform QTL to find out which parts of the genome (and which genes) are most highly associated with your traits of interest. This has been used extensively in agricultural genetics.

After that, my interest would be to get gene expression levels, to learn how genes act in combination.

All this is a huge undertaking that takes a lot of manual intervention, either from an expert or a highly-motivated and highly-talented non-expert. DNA sequencing costs are falling dramatically, but the cost of using that sequence productively is not falling much. This dataset probably cost < $20k to generate, but it's going to cost at least that much in person time before anything useful can come from it.

Plant bioinformatician here. Long story cut short, not much until we get the whole genome built, and a bucketload of plants grown. There's a lot more work on the wet lab side to be done. I'm guessing there aren't the genomic tools in place for this plant to do any funky analysis too. And plant work takes a lot longer to do than other molecular biology. Give it 5 years or so.

BioAcademy (YC W12) is a Codecademy for Bioinformatics.

..coming soon..

It depends on how you want to process it. It's not always that you want to apply complex operations on the whole data set. You may want to process only certain subset of it.

Disclaimer: Biology n00b.

I wonder what sort of structural/symbolic compression[1] techniques will be possible as we understand more of the DNA -> actual stuff process.

[1] I'm thinking of something like Google Courgette, a way to generate symbolic/relocatable diffs for binaries:

http://www.chromium.org/developers/design-documents/software...