 Mae'r bif suicidal sydd y maen o'n gyfr Mae gennym i'n meddwl yn ystyrgynt. Rwy'n canfodol i'n meddwl, a ffanaf i'n meddwl yn ymddangos. Rwy'n meddwl i'n meddwl, wedi gweithio i'r meddwl i'r meddwl am ffanaidd i'r meddwl a'r meddwl i'n meddwl i'r ffanaidd i'r meddwl, bod yn cael ei d sor usandoid y ffordd. Yn gwneud ei gweld i'r gwneud ddweud wedi gweld i'r gwneud ei ddweud i'r meddwl. Roeddwn yn credu cael ei ddweud ei ddweud y ffordd, The rich countries are coming into the steel. So all this invention, the profit from invention, is being made by the rich countries at the expense of the poor. You get this permanent argument that doesn't really have an ending between the farmer fear and developing country together feeling the principle of IP protection is working against them before the developed countries. There isn't really the red solution for that, ac mae'r cyhoeddedd Cd-D yn ymdill ymddill. Mae'r gyflaen oedd gyrraedd yma yn ystod yn gweithio i gweithio eu gwerthodd. Mae'r cyflaen oedd yn 1993 ac yn y gyflaen o'r cyflaen o'r cyflaen o'r gyferllin. Mae'r bwrdd y cwestiynau chyfodol yn y cyd-wyr, yn y pender, yn cyfridd yma, ac mae'r cyfridd ar gyferllun gyfrifau, yn gyflaen o'r cyfridd, yn y cyfridd yn y cyfridd. Mae'r cyfridd yn y cyfridd yn y cyfridd, yn y cyfridd, yn gallu ddweud y brif yn gweithio ynglyn â rhwylo sydd yn cael ei ddweud, ac mae'n ddegos i'r pethau yw'r unol yn dweud hynny fel y dyfodol. Rwy'n ddweud, mae'n ddegos i'r ddweud o'r ddegos i'r ddegos i'r ddegos i'r ddegos i'r brif yn cael ei ddweud. Mae'n ddegos i ddegos i'r ddegos i ddegos i ddegos i ddegos. So, if anybody is worried about the change in the IP philosophy really been very Очень having producing public goods to protecting the IP just think that this new approach to IP protection was exactly what the anti-irrig NGOs want us to be they want us to take control over our IP so that we can prevent mis-use by the deep真的是 And that means well enough Felly arno, mae'n ddodd yn gyhoeddur i'r ysgol yn gwneud o osуys i ddweud, ..wyddech chi'n chlasio'r ffyrdd... .. perfectionist yn fyddfa yn siaradeth. mae'r cyffredin er mwyn i'n ddweud o'r wneud. Byddwch chi'n nhw'n bryd yn fyrdd o'r prysgol. Yn ni'n fryd i'r lleol ar yden, cofio'r sgol yma? Mae'r ymdelineis yma arall eich bod yn ynmysgol yma. Dyna llawer o'r ffordd ag unig o'r gweithio gyda'r golygu, eupon i'r gweithio gyda'r golygu. Ond, mae eisoedd o'r gwir iawn yn ddal, ei fod wedi'u gwasanaeth yn gwneud i'r ffordd. Yr unig o'r gwir iawn yn gwneud ei gweithio siarad. Ytog amser. Ac wedyn ond nad ydym ni'n gweld ar gyfer eu ffordd. Ydyn ni'n fwy o'r ffordd a roedd ni'n ei wneud i, ac ydych chi'n ei ffordd i'r Fygoedd. if you're sending it out, it turns out to allow others to do their own brewyd. That's a little bit more complicated. Depending on whether it's a genbind next session or anything computer-less you receive for the SNTA on the other non-irey line, this is a difficult one. Different kinds of breeding lines, or at least the area of variety. They require different kinds of energy used. The difficult one is this. If you've got non-irey material and you didn't get it with SNTA, Fe wnaeth y gallwch yn uwch i atweithio'r lliwyr arniad wedi eu siam arall, a'r ysgolweld yn gweithio'r cookedwyr, mae'rael i gyfer ychydig yn ddeill sydd honno i chryfodau'r hyffordd fawr o fewn o'u gwirio. Dweud oherwydd y pwynt wedi'n gwirio'r hollol yn ei gofal o Pythoedd. Mae'n ffarnwys yr oed yn amli gymrydol i weithio mor ffarnwys i'w gwirio'r hyffordd. Dwi'n gofyn yn mynd i gydig yn gwirio arno. If you want to take things from farmers, it gets a little bit more difficult. You want to take a germplasm from farmers, you've got to comply with the national legislation on bioprespecting, and that's different in every country. What the CBD requires is government level what they require, what they call private informed and central pick. You've got to get the government to give their own consent before you do this, and it's got to be based on information. You've told them what you want to do it for. In the way of critical, which is the recent extension to this, he encourages you also to get picked from the farmer, and a lot of countries require this as part of the national legislation. The treaty tries to make things easy, it invites all farmers to share this with you, and treaty member countries are obliged to take steps to encourage the farmers to share this with you. These haven't really been awfully effective at the moment, they're still pretty difficult. If you want to get germplasm from the farmer, you'll need to make sure with the national authorities of the country that they're happy with you doing this. Taking traditional knowledge from farmers, the thing that SSD people do, is a lot more difficult. On certain, it's been the subject of pretty intensive discussion, intergovernmental discussion for 20 years or more. I've put it previous here because they don't seem to be much progress. They all say this is really important to do. We have to protect farmers and traditional knowledge, the authorities to know what that means. So move on now, how to conserve that. This is another area where people tell us how to do things, because they think it's really important. We can't be left to our own devices to do it properly. The standards were first defined officially by FLAO as a PGR as it was in those weeks back in 1994. Bystanders have just been endorsed by the FLAO in 2013, but these are all based on committee decisions. We haven't had much scientific input into them, so what we find is that they're not actually adequate. So one of the exciting things we've done over the past few years is to set up a new conservation research unit to address this problem that actually we need to improve the standards. Here's an example of germination rates after a number of years, a varying number of years of storage in the human rights. Of course, most of the germination rates are up here in the 80-100% range where they should be, but actually far too many much lower germination rates in the human rights. We need to sort out why we're getting so much inadequate germination. One of the things we're finding is that it varies with genotype and environment and first harvest processing. Here's an example where rice has been held in temporary storage for either a short time or a long time before we can process it. We have to put it into the cold storage and you see that if it's undergone a long delay, then the germination rates of survival under storage is much worse than if we could process it quickly. Sometimes we're fortunate because a lot of you want germ pleasant for us. We try to make that the priority. If people want seeds urgently, then seeds, like seed-persons, they get put to one side for a while while we're handling your requests. That unfortunately has this knock-on effect that the remaining seeds don't survive with it. There's genetically environmental effects. Here are four varieties being put through accelerated aging tests where we put them in strongly aging conditions for a short period of time. The black lines show seed-producing the tri-season, the red lines see-producing the red-season, the red-season, the red-season, the red-season. For four different varieties we've seen for two debonico times. They're showing a very short lifetime in terms of this house time. So there's genetic variation, there's environmental variation, and somehow we need to sort out this quite complicated trait in order to maximise the longevity of the seeds. We need to scale up capacity because we're expecting much higher use of accessions, genetic stock, particularly in the future. So we're looking at things like automated germination testing, which we are just looking at right now. Automated characterisation that is getting grained and whipped through image analysis, which is almost there. Just like that highlight. Notice you can see the individual hairs there, getting this kind of high-resolution picture becoming critical because when you can make the computer distinguish the actual edge of the grain rather than seeing this as just a fuzzy boundary. So there's a whole load of areas that we're hoping to improve efficiency, which allows us to keep the machine in supply more effectively. Which leads on to the last bit of what I want to talk about, the usage. Everyone says we need to use the collection better. To me, about 85% of our accessions have been screened by area scientists for a long time, or sometimes over 120 times. Last week Bob said it was underused, only 5% were used in reading, and that's true if you look at crosses. I don't know whether you consider this a good nutrition rate from 85% being evaluated to selecting 5% that are considered good enough for crosses. It might be appropriate, or you might show that what this is showing is that we're not able to choose the varieties better, but we'd like to be able to give accessions to breeders to all likelihood that they're used. But I would just like to highlight these gaps. If you're interested in testing some accessions that we've never tested before, we've got 16,000 old accessions that's never been evaluated since 1985. Things have changed since then. Maybe if they were evaluated again now, we'd be able to find them more interesting. There are 12,000 collections, accessions that are collected after CBD that have not been evaluated, which is the majority of them. This is interesting because after the CBD there was concern amongst the breeders. A lot of the breeders decided they didn't want to lose gene bank accessions because of the CBD, and that's reflected in this very high percentage of these closely-reduced accessions that have not been evaluated. Now that we've got the treaty in place, we might like to consider having a look at these things that have now been looked at outside QSC. We've got a small percentage of those treaty accessions that have not been evaluated, that should not have been evaluated, which then have been evaluated. But they're starting to come into the new varieties that we have in the book. Here's just the origin of the old varieties that haven't been evaluated since 1985, but in 90 different countries. Quite a lot from China, no one's looked at the 30 years. Just have a think if you'd like to try. These closely-reduced accessions, a lot from Laos, not so many other countries, and about 20 different countries, but still a range of universities. One thing that's been around in gene bank philosophy for a long time is that we need to characterize them, because that's what makes it useful. You know what they're like. We will use them all. And we have gone to a lot of everything in gene bank to characterize them. These are figures that are a few years out of date, but we've already done the 95, over 90% characterize book. Most of the standard traits, but these tend to be higher heritage traits. We look down here that they're in the same order. But these are not particularly of having a lot of interest to most breeders, until you come down here to these to 80% of the flower. That's an important point, which is up here, is having a lot of data for that. But a lot of this has turned out to be not of much interest to most users of gene bank. So it hasn't been nearly as effective in promoting useful gene bank. They used to say, the second thing we've been told time and time again is that we should publish our data online. Say, here you are, you've got a database for you. You can choose it. You can choose what you want. And users ask us to do that. Other gene bank organisers ask us to do that. But actually, when you look at the kinds of questions we want, you look. These are typical of a real request we've had. But as I said in the countdown I told you, there's so many different variants of these that we've just asked us, but because of that I won't know what to give you. You ask, as I said, I won't know what to give you unless you tell us which particular one. Nice that's good for growing in Canada, or UKs, or Ethiopian highlands. Any of these things, droughts, chief blood, climate change, I can't begin to tell you which is the best accession solution. How would you convert these kinds of requests, think that we just want, into a database search? That's not a trivial exercise. I'm sure it would be worth doing, but we haven't yet found any online search mechanism for any gene bank in the world that actually translates what we know about accessions into what we just do such as we want. Another thing we've been pushing into doing is making call collections. There was a paper a long time ago that claimed that we could get money for something worse with 10% of the collection, which is pretty much a figure for a matter of the year. They were saying that would give you a small amount of collection to evaluate, but still people find it 10% to be. So if our collection is about 12,000, you're not going to evaluate which amount of accession is. So that people have invented mini-cores, micro-cores, and nano-cores. We have call collections over the whole range of 20 up to 12,000 accessions. You're welcome to use any one of them if you like, but why I've put down here, this is a notion of theoretical frequency distribution for value for a defined objective when you're not in particular trade at a location and you're located at a production system. You're wanting to get maximum value. This is a frequency distribution of gene bank accessions. This, almost by definition, is what a reader's collection could look like, because the readers have already gone through picking out the best. If we try to argue that their collection wasn't better on average an hour than the readers would have been able to do. Moreover, they have tested it for their purpose and we haven't tested ours for our purpose. So why on earth should you waste resources on evaluating material that you know has a low average value? And the same question applies to a call collection, to a small subset of collection. What you really want is to get up at these valuable accessions that are better than anything a reader has. But how do you find that? You either need some very exploratory reader, or you need a gene bank into a different way of doing things, a lot of some of the readers. Some old readers who are close to retirement will be much more cautious and say, I'm not going to follow a reader all this useless material. Another strategy is to use GIS methods. If you know where a traditional address came from, you might go to believe something about its properties, something from high up in the pool is getting more co-towered than something on the pangas. So you can, in theory, use GIS approaches to better selection of accessions. But actually, when you look at what we've got, this is an advantage to accessions where we know data on these gets a passport data. So we know the country of origin from what's almost everything. But still, even if you go down to another province, there's only half the collection of your own province again. And when you look for that longitude, there's only about 40% of the collection of that longitude collection of data. So applying GIS technologies is only going to work for various kinds of subsets. So it is not particularly effective in practice for most of the collection. Another one that GMATs have been told to do for many years is to get into pre-breeding, where we cross some clear accessions into elite genoclasm to make them more attractive to readers. I'm thankful that we don't have to do that, because BBGB has been doing that very effectively for many years, bringing in new genoclasms, especially for the purpose of making it easier to other readers. The next one is phenotype. Everyone is saying we need to phenotype. That's the big bottleneck, is understanding of phenotype. But actually, we can't do that in the whole collection. It's not even one train. If you look at the difficulty of looking at the stress tolerance, you might have low heritability, you might need lots of replication, you might need lots of locations, you might need specialist equipment, you might need unstressed and stressed treatments. It's a huge task to leave it for one train, and no one could ever contemplate screening of 120,000 concessions. There's specialists in those stress measurements. You can't do it when you're reading the human access. And it's very inefficient. If you're looking for code points, you're not going to start with concessions on sea level in the tropics. And more importantly, again, the value of concessions actually lies in the phenotype of its programming part of itself. We can't think about what ways we can produce this better than our concessions. It may be okay if you're looking for major genes, but the phenotype of the concessions will really indicate what it's functioning like, but it's really not a good predictor when it's looking at small-effect ptls and genes that have a fantastic interaction. So it's not very effective if you're looking for transgressive segregation, and that's what we need to be looking for when we're wanting to invent new varieties that deal with new problems. We've got to have transgressive segregation. So we really have to phenotype the concessions. But if it's impossible just to phenotype the accessions, it's even more impossible to phenotype all possible, but it's really out of question. Which leads us nearly to the end to get very quickly about the way we're thinking about it. If I went into this in proper detail, this would take another seven hours on its own, and I hope Ken will do that one day for all too long. The DNA has a very active characteristic for the gene-like. That is, 100% heritone, more or less, we can, in theory, measure the known accession, and that's what we need to have an indicator of what an accession might be useful for. We must have data on it. If we don't have data on it, then we've got no idea what an accession would be useful for. And as you know, the one that happens with phenotype. So the idea now is that we develop the training set during the gene discovery by sequencing and phenotyping our core sets of between 23,000 accessions, and they're probably using all sorts of methods, the magic, the nest association of the population, the wide range of the population. The mutants hope to see for mutants that they're trying to discover the function of every gene in the rice genome that will contribute to this and that of the relations between genotype and phenotype. Once you have that for this core subset, you can start in saying, well, we can make predictions about the potential value of the other accession but we haven't. And Bob last week was talking about theory 2035. This is what we invented just theory 2035. We'll have all accessions for you. We won't have all accessions phenotype. We'll never have all accessions phenotype. But if we have this knowledge, we can then, we can combine our idiotite, whether it's a new plant type or whatever the idiotite is of 2035, we can use this knowledge to convert that phenotypic idiotite into a sequence idiotite and design that on the computer and then compare that ideal sequence with what we have in the gene bank and we can construct the best reading system, which accession will be most efficiently to that new sequence that you've designed and we can use it to not only to choose the best of the parents but also to design the best of the markers to use the marker-assisted solutions. And we will only ever phenotype to validate or refine the predictions. Of course this will be cyclic but first predictions aren't going to be very accurate but as we do not have the knowledge about gene type of type of relationship you can gradually get better and better and better. So you will always be phenotyping with only guided phenotype, not just phenotyping anything, phenotyping in order to help us in this time of relationship between genes. Getting to them is a big thing, it's almost like saying being the upscotty is that kind of thing. There's still way in the future, it's not going to happen to long after I've left you in, there's some very big data management and informatics decisions required, we're going to have to come up with normal analyses of the genes in this type of relationship. We have already started putting genetic stocks in the gene bank, and we think we're pretty keen to helping to build up this knowledge of gene type in the time. The reference genome is a big problem, the fact that we've only got little barrier at the moment, we need to be building up much better references that will be useful across the entire rise of gene. That's a very important short term target is to improve our reference. Of course sequencing technologies are going to improve and we need to be both erosion of the time, hopefully reducing the amount of assembly required to work from that point on. We need, in the medium term, to optimise our pipeline to discover it first. We're not going to just say let sequence our collection, we need to find out an efficient way to sequence how we do the first 6,000 and the first 10,000 and the first 20,000 and so on, and optimising the gene fleet type prediction system, just what kind of algorithms do you need to get the predictions? That allows us to do these things. I'd like to acknowledge the GRC team, everyone involved in this, but not only the GRC, everyone else who contributed to Theme 1, not only those who didn't. No one contributed to Theme 1, but a lot of people out there. We had to be part of Theme 1 without knowing it. Anyone who's evaluating and using rice diversity is helping us to get there, and even those who have a view of political and legal stuff, please help us through this. So, empty very much.