[MUSIC] >> The next step. >> In the Benson work, was totally dedicated to topology. The topology is the arrangement of the subunits in the structure. >> It is the internal structure of the gene, that he's dealing with. The Watson and Crick model predicts that the gene is a linear succession of betas. A linear succession of units. Now, of course he talks about, he's conscious about nucleotides, and he talks about nucleotides, but he is also, in that sense he's an absolute follower of Delbruck, he doesn't want to have preconceived idea. And he wants to test models in an unbiased way. So when you make a map of mutant, you will see how distant they are. You will say whether distance can be additive or not totally additive, or partially additive and so on. But, the topology question is the connection between the parts. Topology asks qualitative questions. Not quantitative questions. Now I want to take this first example which is purely imaginary, but where you have four mutants. And these four mutants Intersect. That means that mutant one will not recombine with mutant two, will not recombine with mutant four, but will recombine with mutant three. This is what you see on this slide. Take mutant two. Doesn't recombine with one, doesn't recombine with three and recombines with four. That's exactly what you see here. Three recombines with one but not with two and four. And four recombines with two but not with one and three. And of course the diagonal is full of zeros because the mutant cannot recombine with himself. Now you could arrange this array in these three different ways. You could start by one, two, three, four. You can start by one, two, four, three. You can start by one, three, two, four. You will get the zero and ones mixed in this table. This to me at first, the first time I saw it, looked absurd, because I knew the genes were DNA. But imagine now you have a circular chromosome, like a plasmid. And you start making deletions in this circular chromosome. And you make overlapping deletions. You could get a recombination map that looks like this. So a circular chromosome could give you something like this. It's not totally absurd. So basically what Benzer is setting up is, and you see the physics student appearing. He's talking about zeros and one. This is a language that would become the language of informatics, of computers. So, he takes another example where he has six mutants and he said, well, let's not imagine something as crazy as circular. But let's imagine something simpler as branches. What is the prediction of a branch structure? So he decides to draw one example with six mutants that remove the black portion of these structures. And then, you ask who recombines with whom, one recombines with nobody except six, this is this etcetera, etcetera, etcetera. Four can recombine not with one, with two, with three and with six. Four can recombine with two, with three and with six. This is possible. Now, unless you find example like this, you cannot say that you have it. When you find them you can say you have branches. If you don't find them you can say we haven't seen branches. We cannot say there are no branches. This is a very important notion. The negative result and the positive result don't have the same meaning, or the same strength. So basically what he noticed is that if you have a branch structure, the zeroes are interrupted. Let me just show you this with the red. If I go any, say I take mutant 2. Mutant 2, the zeros are uninterrupted. But if I take mutant 3, I could take the other color. The zeros are interrupted by what? So you have missing, present, missing. You have interruption in the 0 line. Interrupted 0 line. Because you touch a branch. The two that doesn't touch a branch is uninterrupted uninterrupted, OK? So that's very simple. This is an example. So basically what he does is he takes 145 deletions. He takes deletions in this paper because deletions are supposed to remove parts of the gene. So if you remove a part you're in better chance of removing a branch than if you only change one nucleotide. So he takes these deletions and. >> So he makes crosses between this deletion, this set of 19 mutants. And you can immediately see that H 23 cannot recombine with any of the mutant. H 23 is a big deletion. But some of the others are big and some of the others are small. H 23 is big. Now on this drawing, the order of mutants, this order is the order of isolation. On day one, he isolated 184, on day three isolated 215, and so on and so forth. So they're ranked by isolation order. Now, if I go to the supermarket and pick people at random, the first one I pick may be small or large, man or woman, kid or old person, anything. The second one is again random. There's no choice. I just picked a sample of the population. This is what you see here. Now you can rearrange this table into another table, which gives you, this time, uninterrupted rows of zero. There is no series of zero that is interrupted by one. Take, for instance, this line. This line is T782. This is 782, so you go up here, you have 000000, they intersect 00000. Uninterrupted zeros. This one has only five zeros uninterrupted. Et cetera, et cetera. In this example, he found no evidence for a branch structure. And actually, he tested the 145 deletions. Now, that's a large number of crosses, we didn't do all of them, but it's a large number of crosses. And now, again, there's no interrupt. >> Which essentially means that. >> It is the nature of the present analysis that the existence of complex situations cannot be disproved. However, the fact of the matter is that a simple linear model suffices to account for the data. He does not say because a correlation is good, that model is right. Suffices to account for the data. And the last sentence of the paper, these results permit representation of the mutation as alteration in a linear structure in which the functional unit defined by the sys trans test correspond to unique segments. The possibility of branches within the structure is not necessarily excluded. And the branches Max Delbrueck is thanking this paper for suggestion regarding the possibility of branches and for his usual moderating influence. Now anybody who had known Max Delbrueck would laugh at this, because Max Delbrueck was certainly not a moderate. But Delbrueck was very sharp. Benzer was extremely sharp too, and somehow Delbrueck managed to moderate Benzer, which tells a lot about how Benzer was behaving at the time. So in fact now that we know in more detail the structure of the region. The divisions have be mapped and in fact you see that the nineteen divisions are indeed arranged linearly. It was a help that in fact in this case, these four mutations were in fact double isolates of the same event. They behave the same and they are the same event. We isolated over and over and over.