 Okay, we're going to go on with the next presentation for the day, and again, once again, a thank you to our sponsors, Family3DNA, thank you also to Jared Corcoran for live streaming these lectures on Facebook with the permission of the speakers. So it gives me great pleasure to introduce Andrew Millard. Andrew is associate professor of archaeology at Durham University. He has a BA in chemistry, a D-Phil in archaeological science. He's a member of ISOC Society of Genealogists, the Guild of One Name Studies, and a variety of family history societies. He also is the chair of the trustees of GenUKI, which I think you probably pronounce GenUKI. GenUKI. GenUKI. GenUKI. But Andrew also has a keen interest in the Scottish prisoners, and you have a website and a blog about that as well. Andrew Millard. Yes, that's a project. They Scottish prisoners of war. They were in prison in Durham. We excavated some of their skeletons five years ago, so there's a project website that with their descendants of some of those who survived, went to New England. There's a lot of genealogy to do there as well. GenUKI. Yes, I'll turn that up as well. But so today what Andrew's going to talk to us about is the Watto tool. What are the odds which he has provided the statistical input and has worked very closely with Leah Larkin and Johnny Perlman on this particular tool, which is another great tool to use with your DNA matches. So can you please give a big warm round of applause for Andrew Perlman. Andrew Perlman. Okay, so I start by saying thank you to Leah and to Johnny for their contributions. Leah came up with the idea. She said, can we do the maths for this? We sat down and we put it into a really clunky spreadsheet. And then Johnny came in and has created a slick interface that is what you're going to see today. And bizarrely, since we started doing this, Johnny and I have discovered a DNA connection between me and his family. So Watto is about analysing your autosuit or DNA matches. And I'm suspecting that, I suppose I ought to ask, how many people have tried to use Watto? View, okay. So the idea is mostly to help you fit in those matches that have no trees. So it's really useful if you're working with adoptees or unknown parentage of some sort, or where you've done the tests and you've got an MPE. And then, but you don't know, of course, the unknown parent is the person you want to locate. So it's based on autosomal DNA. And I probably don't need to tell much to you. We have the 22 autosomes plus the sex chromosomes. We're not using the sex chromosomes here. You get one copy of each from your parents. And they've got one copy of each from their parents and so on, going back. And that means that you can say what the average sharing is expected to be with various different relationships. So we expect, you should have a point at some time, you should expect to share, well, you will share half of your DNA with your mother and father. You will share approximately half of it with a sibling. You will share a quarter of it, 25% with an uncle or aunt or with a grandchild and so on. And as you go out to more and more distant relationships, the amounts reduce and the averages reduce. But of course, as we all know, when we look at those third, fourth, fifth cousins in our analyses, the amounts are very variable and they would get relatively more and more variable as you go further and further away. But we can still use, this chart is really useful because you can put these relationships into a series of groups which are expected to share the same amount of DNA on average. And that means two things. We only have to think about these groups. If you've got two relationships in the same group, you're never going to be able to tell them apart on the amount of DNA because the sharing is expected to be the same. So the amount you expect to share, say with a first cousin twice removed, will be the same as with the second cousin. And if those are the two relationships you're trying to distinguish between, you can't distinguish between them by the amount of DNA. You have to bring in other things. Of course, the companies will give you ranges. So Ancestry published this chart showing if you share this amount of DNA, they predict these levels of relationship. And it's an interesting chart to look at because you'll see that there are gaps. So what happens if you find between 620 and 680 centimorgans, I have no idea, Ancestry seemed to be not predicting what the relationship is according to this table, but they will predict something. But I've never seen that on Ancestry. I've never seen any removed on Ancestry and most of the companies will stick with simple cousin relationships. They won't talk about half relationships and they won't talk about removed relationships. So I think you'll see those on my heritage. So the companies tried and give you these predictions, but we all know that when we start reconstructing the genealogies we often find that the reality is different from the prediction, particularly as you go to more distant relationships. It's usually, you can usually get this immediate family, grandparents type relationships reasonably well. So because of this problem, Blaine Bettinger came up with the shared CM project where, so this is crowdsourcing data on known relationships, what the companies say the shared centimorgans are. And this has gone through several iterations. And now you can see that actually these ranges are wider. So these ranges are wider than the ones Ancestry published, they're wider than the ones that Family Tree DNA will use. And if you compare a match to the ranges on here you'll see a greater number of possible relationships than you're getting simply out of the company estimates. So that's one way that you can use your sharing with an individual to get a better or a more realistic estimate of the possible relationships, a wider range than any of the companies are going to give you. But it is limited because it's crowdsourced data. So we know, although Blaine has done some filtering, we know that there are probably misstated relationships in there. There are, when there are unknown double relationships then the amount of DNA will be overestimated for the relationship that's been declared. And there's also a bias in that very, whoops, very few of us have data on the very distant relationships which, so they're pushed out here on the side of the chart because there's very little data. But even if I asked you what you share with your fourth cousins, I imagine that most of the fourth cousins you know about the sharing with, you've found them through DNA rather than the other way round. And therefore it's a bias towards the DNA matches that you tend to investigate which are the higher ones. So that tends to bias this chart slightly to the high side of estimates. Ancestry, actually their estimates are supposed to be based on some simulations that they did. They took real genomes of real customers and simulated the recombination that would happen if they produced offspring and did this over many generations so they could see what real genomes look like when you try and match them to one another rather than just a theoretical sharing so that if there are this hidden sharing in the deep ancestry of the population they come from that will be reflected in these totals. And they published this chart in one of their white papers. So you can see what they're saying here is that if we take the red line for example, so this is first equivalent of first cousins, there's if you share something like 1300 centimorgans then it was equally likely to be a first cousin or a half sibling or an unclean relationship. As you go down from 1300 the probability that that is a, that that relationship is more likely to be a first cousin relationship than a closer relationship in this area between 650 and 1300 the most likely relationship if you get a sharing there and only thinking about simple cousins no half relationships no removed then it's likely to be a first cousin relationship. But you can see here the blue line comes in this is the first cousin once removed and the probability is going up so some first cousins once removed will have more than 650 centimorgans and you've got then at this sort of area here to decide between the two relationships. So you think you can say something about the possible relationships and you can say something about the probabilities of those relationships so if you're looking at the shared centimorgan chart you can say yes or no to a range of relationships but you can't say which ones are most likely. With this sort of chart you can say if I'm up here at 900 centimorgans then chances are it's going to be a first cousin relationship rather than the first cousin wants to remove the relationship but I can't throw out the first cousin wants to remove the relationship. At 650 those two relationships are equally likely. So you can use this so we may be able to use that as well. So this is Leo's comparison of the ranges from various different sources so the shared CM project are looking at grouping those relationships by the letters that I showed you earlier. Ancestry DNA looking at the entire range of those lines on the chart I just showed you and the ranges that have been produced by the DNA detectives group. So you see that these are slightly different but broadly comparable so somewhere around 500 to 1300 or 1400 is group C which is your first cousins. The other ranges vary slightly more. But again if you're using these then you can just say yes or no to a group of relationships. So Leo's idea was that we take these ancestry simulations and we can label them with the different groups and we can then start using the probabilities to say something about what the relationships are. And she produced, Leo went through the painful process of digitizing all this so turning the ancestry chart back into numbers so we could have it in a spreadsheet and do calculations from it. And that means you can then go along with your the amount of centimorgans that you match somebody plug it in and see what the different relationships are that it produces. So we had a spreadsheet that did that. It was a bit clunky, it was not very user-friendly but Johnny turned it into something nice. So here if you have a, so this is Johnny's tool which combines the shared centimorgan data and the probabilities from the ancestry simulation so you can see the results of comparing with both of them. So if you enter the amounts of that you share with somebody, so I've put 336 centimorgans in here and it will produce a series of probabilities for the different groups of relationships. And if you, you'll see here that where these are just plain then the shared centimorgan project data and the probabilities both say this is a possible relationship where you get daggers then these, this value is outside of the range that's been seen in the shared centimorgan project but it's possible according to the ancestry simulations and where you get two stars it's the other way around. It's in the shared centimorgan project but it's not possible according to the ancestry simulations. So you get the summary of what both of those charts predict and you get probabilities from the ancestry simulations. And if you scroll down you get a version of the shared centimorgan chart with all the impossible relationships grayed out and just the possible runs highlighted in color. So that's all well and good if you've just got one match and you want to know what the possible relationships are but what if you've got two matches or three matches and you know how they're all related to one another and there's one other person who's unknown and you want to know how they're related. Well you can go through a process of elimination you can run this for each of the matches and see which set of relationships fit the different possibilities of the family tree. But you should also, it should also be possible to take those probabilities and do something more sophisticated think about which are the most likely relationships rather than just the the possible ones. So that was the idea of Watto of what are the odds is to take the probabilities and think about all the possible relationships and tell you which ones are most likely. So this is the the basic Watto interface. You click and you can enter individuals in a tree it's designed to work with the descendants of one individual or a couple and then you can build out the tree of descendants from them. So when you click on each individual you can set their name you can give them a child and get rid of them you can say how much they match the person you're trying to fit into the tree and you can define half relationships with their siblings so and you can say this is a place where I think my unknown person might fit in the tree. So here I've tested Graham and Hetty and they match somebody at 300 centimeters and 236 centimeters where is this person going to be in their tree. This is a sort of second cousin sort of range but other relationships are possible and you can see here on this this side of the tree there are a series of known descendants of and relations of of so descendants of their their ancestors who where this person might fit in the tree and we might ask is Ken the father of our unknown person or is Jack the father of our unknown person. What the tool does is it takes the probability looks at the shared St. Morgan's it looks at the relationship here so hypothesis one would be a second cousin once removed to both of these people hypothesis two would be that a second cousin to both of these people and if we look at the amount that they're sharing then what we find is that the second cousin once removed relationship is ten times less likely than the second cousin relationship now that's a very simple example just two matches and two possible places in the tree you can get build it up to things which are much more complicated so you can add other hypotheses so here extending that tree what if Caroline and David had a sibling that we don't know about and if that sibling had descendants would that be a better place to fit this person in the tree should I be going back to do some more genealogical research to see if this is a real possibility so you can see here that we get again the relative scores for the different hypotheses that our match is a sibling to Caroline and David David is not possible on the shared amount of DNA that is shared so we get a score of zero and a red marker at that position in the tree the other two possibilities are possible we get a green and a score but you can see that they're the these are equally likely as the other the two that we've got already because we really haven't got very much information here with just two matches at that sort of distance so this is another second cousin well you know if one second cousin is possible any second cousin is possible we can't tell those relationships we can't tell one second cousin relationship from another simply on the DNA we would need some other genealogical information a first cousin once removed relationship has about the same probability as a second cousin once removed relationship so they only they're 10 times less likely than the second cousin relationships so if we got to that sort of stage with our tool then what we actually need is more data we haven't really got anywhere to say anything useful so let's test some more people and add them to the tree so if we find Nick and he's a great grandson of David then he's going to be more closely related to these two possible places in the tree than the people we've tested already and that helps us to refine our hypotheses so you can see that hypothesis two becomes impossible if Nick only shares 53 centimorgans with our unknown person 53 is too little for him to simply be a first cousin once removed so that relationship just is eliminated you can see that this is still possible for him to be a second cousin and for these for Graham and Hetty to be second cousins once removed that all works on the but it's unlikely that there's an unknown sibling with a child who is our mystery match is not possible but it's possible there is a grandchild now you have to build your trees you have to think about all the possibilities of different places to put the person in the tree things that are possible genealogically I haven't given you anything any dates here but if you know the ages of some of these people you can eliminate them as potential parents or grandparents and there may be other reasons to eliminate them as well if they're living on the other side of the world so it's not just you haven't just got to think about the DNA you've got to think about what's the genealogy what are the possibilities given the genealogical information as well as the genetic information okay so the hypotheses here are nothing I've shown you get to but we've got a score of one which is always the worst hypothesis that is possible the score of zero for impossible hypotheses and here are only 10 for the next hypothesis if you scroll down you'll see more detail about these hypotheses and so a lot of people look at this and then they come along to the Facebook group which is the support group and post something saying what's going on here I don't understand there's uh they haven't scrolled down don't forget to scroll down there is more information about how the calculations have been done and what they mean so uh here are our five hypotheses they're ranked in the order well they're ranked in the the green ones are ranked in their order of likelihood so the best one is at the top and a description here of something about the statistical probability so the best thing we have here is scoring 10 it's 10 times more likely than the next one which is more likely but not really strong evidence so it says it's most likely but not significantly more likely than the next one this one is possible these ones are not statistically possible so we haven't really come to a conclusion with this amount of data if you scroll down again there's more information and that is showing you all the collated match data and it shows you some so your hypotheses across the top of the table the match is down the side the probabilities drawn from that ancestry chart are given and the relationship that is has been inferred from the tree that you put in and then highlighted in red if there's zero that tells you that this piece of information is actually ruling out that hypothesis that relationship and therefore that hypothesis and then you get the actual odds of the hypotheses compared with the worst possible so it's all relative it all depends on which hypotheses you put in if you don't put in something that's possible it won't do any calculations for it because it depends on what you put in so you have to be putting in the hypotheses you have to be thinking about what's possible what might might be possible so if you go on and do more research and more testing of relatives so maybe we do find this third sibling and we find his descendants and we test two of them and suddenly we're in a realm of bigger matches closer matches so now hypothesis one isn't possible anymore and hypothesis two is still ruled out the possibility of a grandchild of Nathan is still there and I've left out some of the ones that I had before but now the chances are that with these higher scores our unknown person should be fitting into this branch of the tree rather than the the two that we had already so you can do you do this as an iterative process you can add as you do more genealogical research add people and possibilities and as you do more genetic testing you can add people into the tree so now you see we have a score of one for our worst half of the system we have a score of 253 for the best one and the question is then how to interpret those numbers so what I recommend as a baseline for interpreting these is this score system that was derived by Cass and Raftery so if you get a score of one to three they say that's not worth more than a bare mention it's some sort of weak evidence these both these are both possible and one of them is slightly more likely than the other but really you don't want to worry about it three to 20 is positive evidence that's probably where you want to focus your research efforts on clarifying that line finding more testers that are closer in that line 20 to 150 is strong and over 150 is very strong evidence so really you want to be over 20 and really an ideally you'll be over 150 in your score so although these this is a useful scale for thinking about how to interpret the scores the odds ratios we do still remind people that Watto is in beta testing and you will see that on every screen that I put up there's a little slash in the top corner that says beta to remind you that this isn't a fully functional tested tool and we're still learning about whether it whether these odds ratios are working exactly as we think they are we've had a few cases where we know what the relationship is and it doesn't you strongly predicts something else so it's not entirely you can't rely on it entirely and what you should really be using it for is a pointer to where to do more research and more testing because once you've identified which branch somebody is it is likely to be in test somebody as close to them as to that predictive position as possible because the closest matches are the ones to where we can be most certain about what the relationship is if you get a match at 1400 centimorgans then you should be thinking about siblings and that's much more useful information than a first cousin match or a second cousin match and if you want to know exactly where somebody fits in a tree so there are a series of things which I should say also about limitations of Watto first of all we assume that the probabilities for each relationship are independent of one another and there's no crossover between relationships that works as an approximation when the relationships when two of the matches are not very closely related so if their first cousins are more distant then that will work if you have a series of siblings that you've tested and they're matching your unknown person then that will tend to bias the scores that you get out of Watto and you need to be a bit more careful in interpreting them and similarly if you've tested a group of people who are aunts and nephew's and they're matching then those people who are closer than first cousins in a group you have to think carefully about whether the the tool the numbers you get out of the tool are actually to be interpreted in the same way. We don't have probabilities to handle things that people are really interested so we don't have any data on double cousin relationships but the ancestry simulations we started with are simple about simple cousins we don't have data on double cousins although there are ways for where the double cousin relationship is some double cousin relationships could be could be factored in if there are things like three-quarter siblings where there's one father and two mothers who are sisters then certainly we're not handling that at all in this sort of tool we can't have that sort of tree and we know that it probably doesn't work for endogenous populations. Ancestry simulated this with people that they were pretty sure were unrelated in terms of DNA matching. If you're coming from an endogenous population then there will be people who are very distant and unrelated in these sort of terms in their seventh eighth cousins but they're sharing much more DNA than you would expect so it doesn't work in those situations. The other caveat is that ancestry didn't simulate what happens below 40 cents in organs or at least they didn't report it so we've extrapolated the curves to use to allow things below 40 cents in organs to be used but that is an extrapolation a really rough approximation and so we recommend that if you're using Wato at least half of the matches in there should be over 40 so we're not going to be able to make and I don't think even if we have the simulations if most of your matches are below 40 cent in organs we're not going to be able to tell apart the fifth cousins from the sixth cousins on this sort of tool because there's simply too much variation and overlapping the possible values that you get at that level of cousin help. Okay that's how Wato works there is an old version which is actually a more advanced version and I want to say a little bit about how to use this so there is this original version which Johnny produced which doesn't have all the slick trees in it it's just a table that you have to fill in yourself so it's a bit more work and you have to do a lot of the background stuff yourself thinking about the trees but it allows you to do things that you can't do in the tree building version so I want to say a little bit about how you can use this particularly in one situation where you have both maternal and maternal matches to an individual and you want to know or perhaps grand matches through the grandmother and through the grandfather or maybe a great grandfather and great grandmother and you want to know where they fit in the tree because these are two different sides of the family you couldn't draw a tree with a single ancestor that branches and comes down to all the matches so this is the tree that I was working from so you can see there are two couples up here they have a series they have a son and a daughter who marry and have children and grandchildren but they're in there the other children also have descendants who we've tested so we've tested Abby Bert and Charlie we've got some matches we want to know where our unknown person fits in the tree are they a grandchild of this couple or a great grandchild of this and if you looked at any one of these individual matches you couldn't say whether which relation which was the right relationship so this would be a first cousin twice removed that's possible with 497 this would be a third cousin possible with 230 this would be third cousin that's possible so we've been second cousin once removed and so on so we have these two possible places the person will fit in the tree and we need to do a calculation so you have to draw out the tree yourself and you have to work out what all the possible relationships are given the different places they might fit in so hypothesis one here if if our unknown person fitted in the tree at this point then Abby would be a second cousin once removed Bert would be a second cousin once removed and Charlie will be a first cousin once removed if they fit it this point in the tree then we get different set of relationships Abby's third cousin Bert is a third cousin and Charlie is a first cousin twice removed so we have the two different sets of relationships but it's not drawn out on a nice chart for you but I can deal with both sides of the family at once that's the advantage and then I enter the sharing of each of these with the unknown match and press the button and it does a series of it does the calculations and you get out a less friendly output as well but you can see here hypothesis one that it's a grandchild of that couple odd to 479 compared with one for the great grandchild now if we've just been on an eliminating relationships thing that wouldn't have we couldn't have eliminated either those relationships but this does suggest that the grandchild rather than the great grandchild hypothesis is much more likely it's very strong evidence that that is the right relationship and therefore we would go back and look at the genius you look at who we could if there's anyone to test and confirm that with a closer relation so that's the advanced version of the tool you can draw as complicated a tree as you like we've done we've done this with a three way tree so the three founding couples with relationships in different cousins marrying different descendants marrying across the three descendants of the three couples and we've done some calculations on that if you can if you can draw it out and work out what the possible relationship zips are for a hypothesis then you can do the calculations so where might this go in the future while it's in beta still at the moment so we want to be confirming what we've got and and verifying that we are working on more simulations simulations of our own so we're not dependent on what ancestor did some of which is a bit black box and we don't know what they did that would allow us to go below 40 centimorgans I don't know if it's going to help very much but we can certainly produce more reliable numbers below 40 centimorgans and we can think about more relationships as well so we could simulate some of those half relationships double cousins of various sorts and then get probabilities for those relationships ideally we'd like to be able to draw more complex trees but that's partly a technical issue about how we draw trees online it's really easy to draw a tree descending from one couple and the interface is really easy to do if you try and put two couples in there it will get a little bit complicated even for you drawing it let alone for making the software underneath it work well endogamy how are you going to make it work with endogamy people keep asking and I have to say no because I can't we can't easily simulate what's going on in dogamy we might be able to draw on some data so there's there is a collection of data going on for Ashkenazi Jews at the moment by Lara Diamond it may be that we can use some of that data to try and get better figures but they won't be quite the same as the ones that we've got and we also want to think about what happens if you've got those close relatives in there as matches to your unknown person can we correct from that maybe I need to sit down and think about maths more complicated maths now how you work that out and it probably depends also on how distant the relationship is which makes it a bit complicated and then finally just to sum up the places to go so the shared CM project tool is on the DNA painter website under the tools menu Watto is there under the tools menu as well and the table version if you want to understand before you go and use Watto read Leah's blog series on science the heck out of your DNA which describes the whole process with a lot and in a lot of detail takes you through all the steps look at that and if you want to know about what a users group then that's on Facebook is just a group called what are the odds this might not help you so much but we Johnny has discovered that we do have a beer I'm not sure if we're named after the beer or the beer is named after us but it probably I advise you to read to drink it after you've used Watto I'm not great thank you very much yes let's just go back there this tool is going to work better in my father than in me correct well in the sense that just on the previous slide you have the fact that it's simulations in the forties 40 cent morgans yes is is kind of the limit but my dad will have double the amount of DNA on my father's side obviously than I do so in that sense it works better for older people who have tested who were bored maybe in 1920 or 1930s is that a general it's a generalization but it's yeah I guess it's as with all of DNA testing if you've got the older generation you get better data you don't the other thing I forgot to mention is that if you have a parent then the child's data is no use at all so just as Martin was saying earlier if you've got the parent and you're wanting to reconstruct a Lazarus genome then the child got all their DNA from the parents so there's no information there's no new information in the child's DNA and if you put the meaning I tell you will just ignore them if you have a match to a parent in a child you will ignore the child and the other thing of course is more complex trees it would be great if you could actually say for adoptees if you have an adoptees match and you get a matrix of how each of the matches match each other put that into the tool and it could generate a best fit tree for your adoptee and they'd be able to see which where they actually sit in the oval in relation to all of their matches is that on the way well there are you need lots of close matches in that matrix if you haven't got any genealogical information at all then you've got to try and recreate the whole tree from the matrix of matches and that means you need to have a good group of close matches where you can reconstruct the small segment of the tree and say here are some siblings here are some aunts and nephews and I can make this bit of the tree and then I can say here's an aunt here's an nephew their match to this person indicates that they are first cousin to the aunt and first cousin wants to move to the nephew and I can slot them in because there's only one place they'll go and then build out the tree like that so you need you need some really close matches to do that okay but it's I know that's good well the medics already have tools to do that yes okay um questions for Johnny now I'm just a bit wary of the the microjohns here because I'm sorry sorry questions for the question and I see one at the back I'm going to turn this down slightly because it may be a little bit on the loud side and I'll turn you down a little bit as well so I'm coming down to the back gingerly I hope we don't get any screams from the from the loud speakers okay question here thanks Andrew I think it is potentially an amazing tool but you showed at the outset are pretty early on a series of distributions showing the series of normal graphene distributions for the expected uh spread yes DNA between the set of the other one now to me that looks like a series of normal or Gaussian distributions and the theory for resolution of these has been well developed mathematically in areas of astronomy and physics was there any consideration given to looking because basically if you look at those they're all basically the same all you're looking at are differences in the I'm getting a bit mathematical I was both mathematician but that's a long time ago differences between what I call the first and second moment first second and third moments of double distributions and they look to me like you could potentially use the techniques developed in astronomy and physics to get better resolution between those and use tools such as twice squared tests or arrange of other statistical tests to get resolution which would tell you actually you know that is the first cousin and it's resolved with a probability of 98 quite something um have any of those techniques been applied or has consideration been given to given it over to mathematical nerves so if you have a series of individuals where you think they all have the same relationship so you have an unknown but whatever the relationship they have it must all be a group of first cousins or a group of second cousins or a group of third cousins then in principle you could do that and say which of these curves fits the distribution of values in that set of cousins the problem is they're not Gaussian they're they're they're complicated and they're complicated by the they're there are theoretical ideas of what the curves ought to look like but they're done on the basis that you've got absolutely perfect matching so most of the most of the data we've got has a cutoff of seven five or six centimorgans segments below that have been stripped out by the the testing companies the theory doesn't take account of that and it doesn't take account of the fact that there will be small segments which are very old so it would work for for these um distributions that don't come down to zero so probably the second cousins you'll be okay but once you get second cousins once removed there should be some zeros in there in that distribution which you're not going to see not going to be covered by the theoretical curves and that's where I actually have to simulate what the matching companies are doing because you can simulate the you can write down the equations for the inheritance but they don't take account of that stripping out those very low matches so you actually then have to simulate what they're doing which is adds a layer of complication but in principle yes if you've got a group of matches that have the same level of cousinhood you could use that sort of approach I'm not sure if it because we're using the full probability distribution I'm not sure that it gives you any more information but I will think about it great thanks very much other questions yeah we'll have one here from the study panel thank you very much Andrew and Johnny and Leah very interesting tool and it's probably one of my favorite I have I'm using it a lot so I have a question on more kind of observation as well when you're working through example you're showing different hypothesis and how the hypothesis change as you're adding new people and one thing I find is people struggle sometimes is coming up with hypothesis so have you considered including suggestions as possible hypothesis for example if somebody matches it at 500 cent more than there's only certain number of limited relationship you can have so have you considered making suggestions in terms of where the hypothesis could be yes we have we've Johnny and I have been talking about how to do that part of the problem is that if you say you've got 500 cent more than a second cousin relationship is possible there are many different places you can put a second cousin in the tree so you can end up with a very large number of hypotheses if you're not careful all of which are equivalents one another statistically so it's but we need to think about how to do that we can we can certainly generate with the closer the matches you've got the easier it is to do that sort of thing if you had a 60 cent more than match as your best one then there are so many places it could fit in the tree that it probably wouldn't be worth doing we have a question here from Catherine and from Paddy Catherine come forward just to get away from the speaker in case we get a screened out by the speaker thank you Andrew and Morris on the future possibilities slide you have down more simulations and the below 40 cm and more relationships to be able to do that what do you need do you need people to join the shared CM project where do you where do you get your data from it's just I'm going to give you that actually is this running it it is going to be easier for you to speak directly into that okay so um we don't we just like ancestry we don't need real data they simulate it it was all done by simulations in the computer I can simulate what the sharing ought to be given the processor of recombination so and then simulate the stripping out small segments by the company's and predict what the sharing will be for given relationships so we don't need more more data to do that and in fact I more data in the shared 70 Morgan project will probably refine that but it won't get over the biases that I talked about which tend to make it to higher values um the if you were working the other way around and testing all of your cousins what distribution would you get okay uh no you can hold on to that I'll give Paddy this well I've come around here behind the mic so that I don't cause chaos I was a brilliant presentation lots of very interesting ideas a couple of things on your last slide that you might add for future possibilities you said you weren't using the X chromosome I'm working in a case where there's an unknown relationship and they're almost sure two ladies have mothers who they think get the same father but I've no X chromosome match which is about a 100 chance but it's possible so it'd be lovely to use this to show them exactly what the options are thanks suggestion I put in my family tree loads of times but I could never go to get confiled to most sites if you have any thoughts of allowing get come up but that's a technical question if you don't modify get come a little bit to allow the hypothesis to be thrown in yeah because also you want a very simple get come with just the relevant parking tree we don't want you to upload your things with 4 000 people in it if you're a good solver you can select which individuals you want to put out into the get come and the last one if you could go back to the chart where you had cousins on both sides that one and they're going to cousin in the middle group as well so a non-kiss so if you had another line coming down here yeah yeah I there's no reason all right good I guess like that because you made the relationship to the two hypotheses that's fine yeah okay hold on thanks buddy any other questions for Andrew great um there was one question that I had and that was in relation to one of the earlier charts and it was possibly oh let's see now ah yes this is what I want to talk about um the groups here this is I'm not sure which group it is but you can see that the first group has got a range of different possibilities but if you are 90 years old the chances that it's your great great aunt or uncle is probably fairly remote so could you actually take the number of people who have reported the variety of different relationships in the shared sense of walking tool and I know it's biased but could you take um the percentages of uh proportionate first cousin second cousin second cousin once removed and so on and adjust it for age so that you could plug your age into this tool and then it would say okay you've got a 48.6 zero six percent chance that it's in this group here but of this group it's highly or it's more likely that it's going to be a first cousin once removed rather than a great aunt or uncle. Some of that is probably the probability of a relationship just given the ages is probably quite difficult to do um because it's apart from the very close relationships where you can rule out some things and say no no this person is 10 years younger than me they can't be my son um that then working out what the probabilities are I mean how many you'd have to know you have to know about the population the individual comes from and you know what sort of ages they typically get married and have children and what the spread of those those things are which gets a bit complicated and I'd rather leave it to you to look at the tree and say this is a genealogical question rather than a genetic question. Sure great great okay well um listen thanks very much Andrew for our fantastic presentation congratulations to Johnny, Andrew and Leah for producing a wonderful tool which has again moved forward to the whole field of genetic genealogism just that a little bit more and has given us lots to think about. Andrew Millert thank you very much. Is it the first question? Sorry what was the question? You asked the first question? Sorry about um I'm not sure actually.