Saved tsdate runs for SGDP and 1000 Genomes?

[clawsoon]

Hi,

I'm not sure if this is the right place to ask, but perhaps someone can send me in the right direction if it's not.

Does anyone have saved tsdate results for the tsinfer 1000 Genomes and SGDP tree sequences (here and here) that they could share for me to play around with?

I tried to run tsdate myself, but I only have 4GB on my Linux machine and it ran out of memory on chromosome 22.

Thanks.

[awohns]

Hi @clawsoon, thanks for your question. There isn't a publicly available, "tsdated" version of the 1000 Genomes and SGDP tree sequences which you link to. Those tree sequences were created using the original release of tsinfer from this paper. We will soon be releasing the dated tree sequences of the HGDP+1000 Genomes+SGDP+Four archaics+Afanasievo quartet from "A unified genealogy of modern and ancient genomes". If you wanted just the 1000 Genomes or SGDP, you could simplify that tree sequence down to the samples you were interested in.
Hope this is helpful!

[clawsoon]

Perfect timing, @awohns. I'll wait for your uploads, thanks.

[awohns]

Hi @clawsoon, the unified, inferred and dated tree sequences of the Thousand Genomes Project + Human Genome Diversity Project + Simons Genome Diversity project are live on Zenodo! Let me know if you run into any issues with the data, I'm happy to help.
The tree sequences with high coverage ancients should be up soon as well.

[clawsoon]

Hi, @awohns - Great, thanks! Three questions right off the bat:

• What's the difference between the _p and _q tree sequences?
• To triple check, is ts.node(i).time * 25 the correct way to get the age of node i in years?
• Will the dates with ancients be different from/more accurate than the dates without ancients?

Thanks again!

[awohns]

What's the difference between the _p and _q tree sequences?

These are the p (shorter) and q (longer) arms of each chromosome.

To triple check, is ts.node(i).time * 25 the correct way to get the age of node i in years?

Yes that's right. This gives the constrained node age estimate (details in Wohns et al. (2021) Methods). If you want the raw estimates from tsdate (in generations), the mean age estimate is json.loads(ts.node(i).metadata)["mn"] and the variance of the age estimate is json.loads(ts.node(i).metadata)["vr"].

Will the dates with ancients be different from/more accurate than the dates without ancients?

It's hard to say: including the ancients "constrains" variant age estimates if a derived allele is observed in an ancient sample that is older than the tsdate estimate of the node age. On the other hand, we didn't directly date the tree sequence with ancients; we used the age estimates from these tree sequences to build that tree sequence. So overall, date estimates should be similar between the tree sequences with and without ancients.

[clawsoon]

Thanks.