Top models reveal all

Ok, the title is a shameless attempt to boost our visit count, but I did in fact want to talk about two modeling studies today — though of very different kinds. I already mentioned the first in a recent post. I have now got hold of the paper on the genetic modeling of domestication and can talk about it in a bit more detail. ((LATER: Science Daily has now done a longish piece on the paper, with lots of quotes from the main author.))

There’s a conflict in the data on crop domestication, as the authors of a recent PNAS paper see it. ((Allaby RG, Fuller DQ, Brown TA. 2008. The genetic expectations of a protracted model for the origins of domesticated crops. PNAS.)) The conventional scenario divides the transition to agriculture in the Levant during the period of climate change around the Pleistocene-Holocene boundary into three steps: wild gathering, predomestication cultivation, and fixation of the domestication syndrome. Based on archaeology, field experiments and climatic considerations, each of these steps was thought to be fairly short: the transition was very rapid. The main genetic consequence of such a scenario should be monophyletic crops. And if you look at the number of mutant alleles connected with each bit of the domestication syndrome (the brittle rachis, for example), or the extent of narrowing of genetic diversity from wild relative to cultivated crop, or phylogenetic relationships based on molecular markers of different kinds, you do in fact get evidence that domestication happened only once in many crops, in a fairly restricted area (barley being an exception).

The problem is that archaeologists have now put a spanner in the works. They’ve changed their mind on the timescale, and in a pretty spectacular way. Rather than maybe 2,000 years, they are now saying the whole domestication process took closer to 12,000 years in the Levant, elongating each of its component stages quite considerably. This extended timeline means that the likelihood of independent, multiple, geographically dispersed domestications of a given crop — and indeed of the different crops making up the Neolithic Package — is much greater. That, however, would be expected to lead to polyphyletic crops.

So how do you reconcile a protracted domestication process with crops which genome-wide surveys suggest are monophyletic? Well, according to the authors, there is in fact nothing to reconcile.

They build an in silico model consisting of virtual plants with chromosomes carrying lots of biallelic markers, put them through one or more domestication bottlenecks and a subsequent expansion, with  varying possibilities for population amalgamation in the multiple domestication case, left the populations to cycle through a range of different numbers of generations, and then looked at the phylogenies for each chromosome.

The result was surprising, even counterintuitive. For multilocus systems, “multiple-origin crops are actually more likely to result in monophyly than single-origin ones.” All the simulations eventually led to monophyletic crops, the speed with which they did so depending on population size: by 2N generations, a crop was monophyletic whether it had been domesticated only once or multiple times.

What does it mean if the transition to agriculture was indeed as protracted as the archaeological evidence suggests — and as the genetic evidence can also be interpreted to suggest, at least based on this modeling study? Well, I suppose one thing that could be said is that the balance between artificial and natural selection may not have shifted as completely and suddenly as was thought. Which would perhaps strengthen the hand of those looking for ways of facilitating the use of large collections through provenance data.

The other kind of model I want to discuss is “climate envelopes.” We have also blogged about this before. The idea behind these things is simple. You dot-map the present distribution of a species. You then extract the climatic data for the places where the species has been observed. That’s your envelope for the species. You then say, ok, what’s going to happen in these places under climate change? Some places will change so much they will move out of the envelope. Other places which are nearby geographically but currently outside the envelope will move into it. The assumption is that, given no change in adaptation, the species will either migrate from the old to the new places, or go extinct. Apocalyptic estimates of possible extinctions due to climate change have been reached using these methods.

But it seems there may be some problems with such an approach, according to another PNAS paper ((Beale, C. M., Lennon, J. J. & Gimona, A. 2008. PNAS.)) We’ve always known that they omit things that are important in determining species distributions: soils, competitive effects, human interference. But there may be an even more fundamental flaw. The authors built climate envelopes for 100 European bird species both based on real data about species occurrence and also based on random collections of points “designed to mimic the spatial structure of the birds’ real distribution.”

The result?

For 68 of the 100 species, the five distributions that fitted their climate envelopes best were null distributions. So climate envelopes generated from real distribution data did not describe that data as well as some of the climate envelopes fitted to distribution data made up without any thought of climate.

Climate is no better than chance as a way of describing the distribution of many species. At least of European birds. Time to test if it is the same for crop wild relatives, say?

Harlan II, day 1

From our man on the spot at Davis, Robert Hijmans.

It is only one day old, but the Harlan II symposium is the best I have been to for ages. That the subject is of some importance helps, of course. In his keynote speech, Jared Diamond called it nothing less than the most important event in the last million year of human history. Guess what, it has something to do with agrobiodiversity. It is plant and animal domestication, of course. Have a look at the program and you may understand that I am challenged to summarize the proceedings. But here are some impressions.

Domestication took a long time. Dorian Fuller summarized archaeological data to show that traits associated with domestication, such as non-shattering of grains, evolved slowly, over 1000s of years. Some speakers distinguished the initiation of cultivation from domestication. I had always thought of these two things as happening at the same time. But why not cultivate wild wheats, or rice? Benjamin Kilian showed data suggesting that wild einkorn was cultivated in Turkish parts of the Fertile Crescent. And Susan McCouch of Cornell University pointed out that, after 4000 years of cultivation, the common rice of West Africa, Oryza glaberrima is not domesticated yet: it still shatters. And I think there are many animal species that are not domesticated but that are nevertheless put to good use. Vicuña for example.

The question whether we domesticated plants or they domesticated us was not (yet) discussed, but there was reference to the self-domestication of dogs and cats. Robert Wayne showed that dogs were domesticated from Mediterranean grey wolves. But wolves are not very friendly to humans, how would you go about taming them? Wayne thinks that it was the wolves who approached our ancestors because they liked to eat the leftovers of their hunting parties. Over time, they may have lost some of their fear and aggression towards humans (obviously not realizing they would end up as chihuahuas). Likewise, cats may have approached ancient towns to catch some of the abundant mice in the granaries of the agricultural revolution.

Molecular biology rules. Remarkable progress is being made in analyzing the genetic make up of crops, the remains of ancient crops, and of crop wild relatives, to solve the puzzles of how our crops and domestic animals left their wild states. In some cases, this work leads to truly new insights in otherwise uncharted territory. In other cases, the molecular work confirms or refines insights that others had obtained from morphological, geographical, and archaeological data.

Even religion was invoked. John Burke explained that sunflower become a popular source of oil in Russia because it could be used during the Lent season, whereas all other sources of oil were on the black list of the Russian church. Gila Kahila Bar-Gal has put the insights from her ancient DNA work on archaeological remains of caprins (goat like creatures) to good use: she showed that the Dead Sea Scrolls were written on goat skin (not sheep). Unfortunately, there was also pieces of (wild) oryx skin — which is not kosher — but these were only used as wrapper; a relief.

Feasting it up in the Neolithic

A Guardian article on the evidence for large-scale feasting at Stonehenge, and in particular on the long-range movement of cattle to the site, reminded me that I had wanted to link to a more general paper about the phenomenon of Neolithic feasting. I have only had access to the abstract so far, but the paper seems to argue that feasting and agriculture went hand in hand, and that in fact the practice may have led to the domestication of cattle. Bit of a chicken-and-egg problem there, at first sight, but I’ll wait for the full text before commenting on that at any greater length. In any case, it seems that barbecues go back much further than the Neolithic.

Actually, I may as well put another marker down. Dienekes’ Anthropology Blog, my source for the feasting paper, also recently had a post about crop domestication. Again, I don’t have the full text yet, and will discuss this more fully when I do. But it seems the paper argues that there is a tension in the data on crop domestication between archaeology, which shows that the process was slow, stop-start and dispersed geographically, and the genetics, “suggesting that domestication (sic) plants are monophyletic, the result of a single domestication event in a definite place.” Well, I don’t think the genetics is saying that at all for many crops, but, be that as it may, the paper apparently presents a simulation model which shows that “multiple-origin crops are actually more likely to result in monophyly than single-origin ones.”

Nibbles: Yeast, Weeds, Bioprospecting, Iraq, Pine wilt, Vietnam, GM, GM, Insects, Bees, Sheep, Fowl

Lost in genebank database hell

Navigating around germplasm databases can be a frustrating experience. A posting on the CropWildRelativesGroup alerted me to a Science Daily piece on tomato genomics which mentioned the wild relative Lycopersicon pennellii (or Solanum pennellii, but I’m not going there, at least not today). But how many accessions of this species are conserved ex situ? And where is it found in the wild?

Ok, so SINGER first, as that’s been much on my mind — and on this blog — of late. SINGER shows 61 accessions of L. pennellii, all from the AVRDC collection. Most of them are from Peru, although 7 accessions have USA, Mexico, Poland (?) or “unknown” as source country. None of these accessions seem to have geo-references, so no nice map from SINGER this time. Pity. But SINGER does give very neat summaries for your query results. ((Incidentally, AVRDC has its own Vegetable Genetic Resources Information System online, which has 65 records for L. pennellii.))

GRIN returns 51 accessions. I can’t find any easy way of working out the duplication between these and the AVRDC material, but I imagine it is significant. Again, most of the accessions are from Peru, but it’s kind of difficult to get summary information across all accessions in GRIN at the moment, though I know they are working on this. Now, tomato germplasm is conserved at the C.M. Rick Tomato Genetic Resources Center (GRIN tells you so), and they have a database of their own. Querying it results in 45 hits, but again there’s no easy way I can see of looking at summary information across all these. You have to look at each individual accession in turn to find out where they’re from, and if you do you get a little map too. The thing I don’t quite understand is why the accessions are geo-referenced in the Tomato Genetic Resources Center database, but not in GRIN. Maybe they’re upgrading the data gradually at the Centre and haven’t passed the latest version on to GRIN? That may also explain the discrepancy in accession numbers. It looks like they’re working on the geo-spatial part of the database, and it may well be possible to get a map of all the accessions of a particular species eventually.

You can of course do that in GBIF right now, but GBIF only has 8 geo-referenced L. pennellii records: from the Missouri Botanical Garden, the Dutch genebank and the European germplasm database, EURISCO. Too bad the Tomato Genetic Resources Center is not a GBIF data provider. And, indeed, that its geo-reference data is not included in GRIN, which is a GBIF provider.

So the answers to the questions I started with are: at least, and probably not much more than, 112, but that probably includes duplicates; and Peru. But I cannot produce a decent map of the distribution of L. pannellii online. I would have to mess around and download the data from the Tomato Genetic Resources Centre database, and then map it myself. Which I may well do, just to show it can be done. But this little exercise does show that there’s a lot of work to be done to improve the data in — and fully integrate — existing agrobiodiversity databases.