WEBVTT 00:00:00.000 --> 00:00:00.610 align:middle line:90% 00:00:00.610 --> 00:00:02.440 align:middle line:84% The following content is provided 00:00:02.440 --> 00:00:06.370 align:middle line:84% by MIT OpenCourseWare under a Creative Commons license. 00:00:06.370 --> 00:00:08.470 align:middle line:84% Additional information about our license, 00:00:08.470 --> 00:00:10.870 align:middle line:84% and MIT OpenCourseWare in general, 00:00:10.870 --> 00:00:13.355 align:middle line:90% is available at MIT.edu. 00:00:13.355 --> 00:00:21.230 align:middle line:90% 00:00:21.230 --> 00:00:22.940 align:middle line:90% PROFESSOR: OK. 00:00:22.940 --> 00:00:26.220 align:middle line:90% Welcome back to the second half. 00:00:26.220 --> 00:00:30.080 align:middle line:84% We'll pick up on this somewhat pessimistic note 00:00:30.080 --> 00:00:36.860 align:middle line:84% of how one goes from a very accurate sequence to a probably 00:00:36.860 --> 00:00:40.430 align:middle line:84% less than accurate speculative league and specificity, 00:00:40.430 --> 00:00:43.910 align:middle line:84% to actually how we actually do get a 3D structure league 00:00:43.910 --> 00:00:46.310 align:middle line:84% and specificity, and some of the powerful methods 00:00:46.310 --> 00:00:52.220 align:middle line:84% that are shared by the computational tools. 00:00:52.220 --> 00:00:55.850 align:middle line:84% One of the things we want to do is we want to, if possible, 00:00:55.850 --> 00:00:57.500 align:middle line:90% find a homologue. 00:00:57.500 --> 00:01:00.050 align:middle line:84% It may be a distant homologue or a very close one. 00:01:00.050 --> 00:01:02.660 align:middle line:84% If it's very close, as we saw from that previous slide about 00:01:02.660 --> 00:01:08.210 align:middle line:84% the highest accuracy coming from homologues that are 80% or 90%. 00:01:08.210 --> 00:01:11.720 align:middle line:84% As we get further and further away, remember we had to-- 00:01:11.720 --> 00:01:13.820 align:middle line:84% first, you could use exact matches. 00:01:13.820 --> 00:01:16.760 align:middle line:84% And then you had to use dynamic programming, and then 00:01:16.760 --> 00:01:18.200 align:middle line:90% hidden Markov models. 00:01:18.200 --> 00:01:20.660 align:middle line:84% And finally, at the bottom of this Slide 37, 00:01:20.660 --> 00:01:22.940 align:middle line:84% we go to-- we resort to threading, 00:01:22.940 --> 00:01:25.850 align:middle line:84% where one of the structures is searched 00:01:25.850 --> 00:01:28.370 align:middle line:84% through a database of three-dimensional structures 00:01:28.370 --> 00:01:30.530 align:middle line:90% with your favorite sequence. 00:01:30.530 --> 00:01:34.100 align:middle line:84% And you not only search through the database of structures, 00:01:34.100 --> 00:01:35.870 align:middle line:84% but through each structure, you think 00:01:35.870 --> 00:01:38.120 align:middle line:84% of every way you could thread your sequence 00:01:38.120 --> 00:01:40.310 align:middle line:84% into that complicated 3D structure. 00:01:40.310 --> 00:01:42.920 align:middle line:84% And some of the threading positions 00:01:42.920 --> 00:01:46.010 align:middle line:84% with insertions, deletions, and offsets 00:01:46.010 --> 00:01:47.280 align:middle line:90% will be better than others. 00:01:47.280 --> 00:01:48.890 align:middle line:84% They will fit that three-dimensional structure. 00:01:48.890 --> 00:01:51.620 align:middle line:84% Some will cause clashes, where the three-dimensional structure 00:01:51.620 --> 00:01:54.440 align:middle line:84% you're searching through and the particular thread 00:01:54.440 --> 00:01:56.840 align:middle line:84% position you've got causes two bulky groups 00:01:56.840 --> 00:01:58.760 align:middle line:84% to lie on top of each other in space 00:01:58.760 --> 00:02:01.543 align:middle line:84% in a way that's hard to relieve the stress. 00:02:01.543 --> 00:02:02.960 align:middle line:84% And so, that threading is probably 00:02:02.960 --> 00:02:05.600 align:middle line:84% the ultimate in getting very distant relationships. 00:02:05.600 --> 00:02:07.670 align:middle line:84% It is limited by the fact that it really 00:02:07.670 --> 00:02:09.282 align:middle line:84% only works if one of the two proteins 00:02:09.282 --> 00:02:10.699 align:middle line:84% has a three-dimensional structure. 00:02:10.699 --> 00:02:13.700 align:middle line:84% But you can search the database of sequences against a database 00:02:13.700 --> 00:02:15.200 align:middle line:90% of 3D structures. 00:02:15.200 --> 00:02:18.110 align:middle line:90% 00:02:18.110 --> 00:02:22.160 align:middle line:84% The antidote to the limitation of having not enough 00:02:22.160 --> 00:02:24.320 align:middle line:84% three-dimensional structures is to launch 00:02:24.320 --> 00:02:26.540 align:middle line:84% a project, like the Genome Project, 00:02:26.540 --> 00:02:30.200 align:middle line:84% to get all of the three-dimensional structures. 00:02:30.200 --> 00:02:35.520 align:middle line:90% Now, is that finite? 00:02:35.520 --> 00:02:37.590 align:middle line:84% Well, certainly for a particular genome, 00:02:37.590 --> 00:02:41.610 align:middle line:84% that number is less than or equal to the number of proteins 00:02:41.610 --> 00:02:45.280 align:middle line:84% in that proteome for that particular organism. 00:02:45.280 --> 00:02:47.070 align:middle line:84% But if we look at organisms as a whole, 00:02:47.070 --> 00:02:49.830 align:middle line:84% where we don't even really even know how many organisms there 00:02:49.830 --> 00:02:53.850 align:middle line:84% are on Earth, then that number may be larger. 00:02:53.850 --> 00:02:56.550 align:middle line:84% But some people estimate that it's less than 10,000. 00:02:56.550 --> 00:03:00.420 align:middle line:84% Basic folds, where a fold is the scaffolding-- which, 00:03:00.420 --> 00:03:02.430 align:middle line:84% once you have that fold, whether it's 00:03:02.430 --> 00:03:04.680 align:middle line:84% a certain number of alpha helices, 00:03:04.680 --> 00:03:06.570 align:middle line:84% betas and turns in a particular order, 00:03:06.570 --> 00:03:08.040 align:middle line:90% in a particular geometry. 00:03:08.040 --> 00:03:11.370 align:middle line:84% Once you have that, and you have any amino acid sequences 00:03:11.370 --> 00:03:15.510 align:middle line:84% within a 35% amino acid identity-- and you'll see, 00:03:15.510 --> 00:03:18.720 align:middle line:84% as this last section of this lecture 00:03:18.720 --> 00:03:21.360 align:middle line:90% goes along, why it's around 35%. 00:03:21.360 --> 00:03:24.090 align:middle line:90% 00:03:24.090 --> 00:03:29.220 align:middle line:84% The goal is to get, saturate the three-dimensional structure 00:03:29.220 --> 00:03:31.410 align:middle line:90% space, so that every three-- 00:03:31.410 --> 00:03:34.170 align:middle line:84% every sequence is within at least 35% 00:03:34.170 --> 00:03:35.740 align:middle line:90% of one of those structures. 00:03:35.740 --> 00:03:38.520 align:middle line:84% This is somewhat conjectural, unlike the Human Genome 00:03:38.520 --> 00:03:43.230 align:middle line:84% Project, where we knew we had 3 billion bases to sequence. 00:03:43.230 --> 00:03:46.890 align:middle line:84% Here, we hope that we have 10,000 basic folds. 00:03:46.890 --> 00:03:50.820 align:middle line:84% And we hope that 35% amino acid sequence identity will be 00:03:50.820 --> 00:03:53.710 align:middle line:90% enough to do homology modeling. 00:03:53.710 --> 00:03:54.210 align:middle line:90% OK? 00:03:54.210 --> 00:03:54.840 align:middle line:90% It's not. 00:03:54.840 --> 00:03:56.760 align:middle line:90% It's not currently. 00:03:56.760 --> 00:04:00.690 align:middle line:84% And the criteria for this has to be prioritized in some way. 00:04:00.690 --> 00:04:03.960 align:middle line:84% And remember, we had prioritization for drug targets 00:04:03.960 --> 00:04:05.680 align:middle line:90% in the previous slide. 00:04:05.680 --> 00:04:08.850 align:middle line:84% And here, prioritization for structural genomics is similar. 00:04:08.850 --> 00:04:11.625 align:middle line:84% But in addition, you want to have-- 00:04:11.625 --> 00:04:14.100 align:middle line:84% you want them to represent the largest 00:04:14.100 --> 00:04:17.250 align:middle line:84% family you can get, but not have it previously solved. 00:04:17.250 --> 00:04:22.830 align:middle line:90% 00:04:22.830 --> 00:04:26.400 align:middle line:84% And for some reason, they're excluding non-- 00:04:26.400 --> 00:04:28.510 align:middle line:84% or they're excluding transmembrane proteins. 00:04:28.510 --> 00:04:30.720 align:middle line:84% Now, this is a very important class, 00:04:30.720 --> 00:04:34.200 align:middle line:90% as we'll see in the next slide. 00:04:34.200 --> 00:04:38.550 align:middle line:84% Because the goal stated up at the top of this, 00:04:38.550 --> 00:04:42.180 align:middle line:84% Slide 38, of assigning functions and interpreting 00:04:42.180 --> 00:04:45.390 align:middle line:84% disease related polymorphisms and drug targets 00:04:45.390 --> 00:04:50.440 align:middle line:84% and so on certainly apply to membrane proteins as well. 00:04:50.440 --> 00:04:53.160 align:middle line:84% And there are reasons that we've already alluded to 00:04:53.160 --> 00:04:55.320 align:middle line:84% for looking specific for programming 00:04:55.320 --> 00:04:56.710 align:middle line:90% cells via membrane proteins. 00:04:56.710 --> 00:04:58.845 align:middle line:84% This is where cell-cell interactions occur. 00:04:58.845 --> 00:05:02.760 align:middle line:84% This is where adhesion motility immune recognition occurs. 00:05:02.760 --> 00:05:05.700 align:middle line:84% These all occur without getting inside of the cell. 00:05:05.700 --> 00:05:11.160 align:middle line:90% 00:05:11.160 --> 00:05:15.200 align:middle line:84% This is a major class of drug targets. 00:05:15.200 --> 00:05:19.670 align:middle line:84% And furthermore, it's not like this is an impossible class 00:05:19.670 --> 00:05:20.930 align:middle line:90% of proteins to solve. 00:05:20.930 --> 00:05:24.530 align:middle line:84% Actually, the three-dimensional structural databases-- 00:05:24.530 --> 00:05:26.120 align:middle line:90% more about that in a moment-- 00:05:26.120 --> 00:05:27.690 align:middle line:90% are filled with these things. 00:05:27.690 --> 00:05:31.040 align:middle line:84% They're certainly-- are the most underrepresented class, 00:05:31.040 --> 00:05:32.630 align:middle line:84% but there are plenty of examples. 00:05:32.630 --> 00:05:34.920 align:middle line:90% And there are two major classes. 00:05:34.920 --> 00:05:38.510 align:middle line:84% One of them is soluble fragments of fibrous or membrane protein. 00:05:38.510 --> 00:05:41.180 align:middle line:84% Some of the fibrous proteins are excluded as well. 00:05:41.180 --> 00:05:43.820 align:middle line:84% Here you'll use a protease to cleave off, possibly, 00:05:43.820 --> 00:05:47.030 align:middle line:84% a tiny piece that makes it insoluble, 00:05:47.030 --> 00:05:48.710 align:middle line:84% maybe a little anchor into the membrane. 00:05:48.710 --> 00:05:51.127 align:middle line:84% And all the rest of it now behaves like a soluble protein. 00:05:51.127 --> 00:05:54.260 align:middle line:84% And we know how to solve soluble proteins. 00:05:54.260 --> 00:05:57.560 align:middle line:84% Other times, the other class is integral membrane proteins, 00:05:57.560 --> 00:06:03.500 align:middle line:84% where they go like this one bacteria 00:06:03.500 --> 00:06:05.370 align:middle line:84% on the right-hand side of the image, 00:06:05.370 --> 00:06:07.010 align:middle line:84% you can see the red alpha helices 00:06:07.010 --> 00:06:11.240 align:middle line:84% go back and forth across the membrane where the gray lipids. 00:06:11.240 --> 00:06:14.060 align:middle line:84% And this is-- there's no way you could clip off 00:06:14.060 --> 00:06:16.310 align:middle line:84% a little piece of this and have a major fraction of it 00:06:16.310 --> 00:06:16.940 align:middle line:90% left to solve. 00:06:16.940 --> 00:06:18.722 align:middle line:84% But this was solved in the membrane. 00:06:18.722 --> 00:06:20.930 align:middle line:84% And you can see the little blue water molecules going 00:06:20.930 --> 00:06:24.800 align:middle line:90% through that channel in this. 00:06:24.800 --> 00:06:27.320 align:middle line:84% That channel is responsible for a proton 00:06:27.320 --> 00:06:31.213 align:middle line:84% pumping, which can be part of the ATP production process. 00:06:31.213 --> 00:06:32.630 align:middle line:84% But there are many other classes-- 00:06:32.630 --> 00:06:39.170 align:middle line:84% redox proteins, toxins, ion channels, photosynthesis 00:06:39.170 --> 00:06:43.340 align:middle line:84% and phototransduction, and so on. 00:06:43.340 --> 00:06:46.280 align:middle line:84% The G protein coupled-receptor class 00:06:46.280 --> 00:06:48.080 align:middle line:84% is a particularly important drug target. 00:06:48.080 --> 00:06:53.780 align:middle line:84% ABC proteins and transporters have also been solved. 00:06:53.780 --> 00:06:57.860 align:middle line:84% So given that this is an underrepresented class, 00:06:57.860 --> 00:07:05.090 align:middle line:84% and given that the structural genomics project will not 00:07:05.090 --> 00:07:09.680 align:middle line:84% necessarily target these in a rapid manner, 00:07:09.680 --> 00:07:11.180 align:middle line:84% what is the current state of affairs 00:07:11.180 --> 00:07:16.430 align:middle line:84% for computational prediction of the transmembrane regions 00:07:16.430 --> 00:07:18.080 align:middle line:90% of proteins? 00:07:18.080 --> 00:07:21.050 align:middle line:84% And actually, I would say the prospects here 00:07:21.050 --> 00:07:22.910 align:middle line:84% are fairly favorable compared to some 00:07:22.910 --> 00:07:25.940 align:middle line:84% of the least favorable ones in that slide, that very 00:07:25.940 --> 00:07:27.740 align:middle line:90% pessimistic slide a few back. 00:07:27.740 --> 00:07:33.110 align:middle line:84% Here, you can get, as indicated in this JMP paper, 00:07:33.110 --> 00:07:35.720 align:middle line:84% transmembrane helices identified-- 00:07:35.720 --> 00:07:38.570 align:middle line:84% and there can also be transmembrane beta, as well-- 00:07:38.570 --> 00:07:42.230 align:middle line:84% but helices identify with accuracy greater than 99%. 00:07:42.230 --> 00:07:46.100 align:middle line:84% And remember, this is basically saying that you correctly 00:07:46.100 --> 00:07:47.280 align:middle line:90% identified those. 00:07:47.280 --> 00:07:49.250 align:middle line:84% And then you also have false predictions. 00:07:49.250 --> 00:07:55.070 align:middle line:84% There'll be a number of peptide regions 00:07:55.070 --> 00:07:57.680 align:middle line:84% of known proteins, which are incorrectly predicted 00:07:57.680 --> 00:07:58.790 align:middle line:90% to be in transmembrane. 00:07:58.790 --> 00:08:03.710 align:middle line:84% And this is a tolerable false prediction rate of 17% to 43%, 00:08:03.710 --> 00:08:08.150 align:middle line:84% given a set of soluble proteins as a negative learning set. 00:08:08.150 --> 00:08:14.730 align:middle line:90% 00:08:14.730 --> 00:08:17.670 align:middle line:84% Now, that's-- just merely knowing that a particular 00:08:17.670 --> 00:08:20.400 align:middle line:84% segment of protein is transmembrane is a big step 00:08:20.400 --> 00:08:22.330 align:middle line:84% in terms of identifying its function. 00:08:22.330 --> 00:08:26.790 align:middle line:84% But to get the further functional characterization, 00:08:26.790 --> 00:08:28.330 align:middle line:84% we need things like ligand binding, 00:08:28.330 --> 00:08:29.760 align:middle line:90% which we've already addressed. 00:08:29.760 --> 00:08:31.950 align:middle line:84% And if you look at some of these quotes, 00:08:31.950 --> 00:08:33.929 align:middle line:84% you can see that a lot of the emphasis 00:08:33.929 --> 00:08:42.030 align:middle line:84% is on display and cataloging and a hopeful expectation 00:08:42.030 --> 00:08:47.670 align:middle line:84% that we'll be able to move from rough three-dimensional 00:08:47.670 --> 00:08:52.470 align:middle line:84% structures to ligand-binding specificity. 00:08:52.470 --> 00:08:54.630 align:middle line:84% But where do the three-dimensional structures 00:08:54.630 --> 00:08:57.030 align:middle line:84% that we do have, do believe, that do tell us 00:08:57.030 --> 00:09:00.130 align:middle line:84% about the exact geometry of ligand binding come from? 00:09:00.130 --> 00:09:01.500 align:middle line:90% Where do they come from? 00:09:01.500 --> 00:09:03.420 align:middle line:90% And how do we compute on them? 00:09:03.420 --> 00:09:04.525 align:middle line:90% How do we read them? 00:09:04.525 --> 00:09:06.150 align:middle line:84% How do our computer programs read them? 00:09:06.150 --> 00:09:11.820 align:middle line:84% Well, this is a typical file of a three-dimensional structure. 00:09:11.820 --> 00:09:13.417 align:middle line:84% This happens to be one that we will 00:09:13.417 --> 00:09:15.330 align:middle line:84% show at the-- the three-dimensional structure 00:09:15.330 --> 00:09:17.070 align:middle line:90% at the very end of this talk. 00:09:17.070 --> 00:09:22.270 align:middle line:84% It is the human estrogen receptor. 00:09:22.270 --> 00:09:24.600 align:middle line:84% And you can see, the first line is 00:09:24.600 --> 00:09:29.340 align:middle line:84% that it is a complex between a protein and a DNA molecule. 00:09:29.340 --> 00:09:32.760 align:middle line:84% And the molecules, the estrogen receptor. 00:09:32.760 --> 00:09:35.620 align:middle line:84% The third line down is the resolution. 00:09:35.620 --> 00:09:39.420 align:middle line:84% This is a technical description of the X-ray diffraction 00:09:39.420 --> 00:09:40.920 align:middle line:90% pattern. 00:09:40.920 --> 00:09:44.880 align:middle line:84% It gives you an upper limit to how precise it is. 00:09:44.880 --> 00:09:47.460 align:middle line:84% It's going to be more precise than 2.4 angstroms, 00:09:47.460 --> 00:09:49.650 align:middle line:84% depending on how much statistical oversampling 00:09:49.650 --> 00:09:53.220 align:middle line:84% you have, and how good your computer program is enforcing 00:09:53.220 --> 00:09:54.900 align:middle line:90% their chemical constraints. 00:09:54.900 --> 00:09:58.650 align:middle line:84% A typical precision for a 2.4 angstrom protein structure 00:09:58.650 --> 00:10:01.890 align:middle line:84% might be on the order of 0.3 angstroms, 00:10:01.890 --> 00:10:05.070 align:middle line:84% maybe eight times better than the nominal resolution. 00:10:05.070 --> 00:10:09.180 align:middle line:84% But that's an important number, unambiguously determined 00:10:09.180 --> 00:10:10.513 align:middle line:90% in the process of collecting it. 00:10:10.513 --> 00:10:11.972 align:middle line:84% So when you look at the literature, 00:10:11.972 --> 00:10:13.050 align:middle line:90% look at this number. 00:10:13.050 --> 00:10:14.620 align:middle line:84% And look at the next number down, 00:10:14.620 --> 00:10:18.150 align:middle line:84% which is the R value, which is not a measure of resolution, 00:10:18.150 --> 00:10:22.150 align:middle line:84% but a measure of goodness of fit between the model. 00:10:22.150 --> 00:10:25.170 align:middle line:84% The model is the X, Y, Z-coordinates of your atoms. 00:10:25.170 --> 00:10:28.840 align:middle line:84% It's the goodness of fit between the model and the data. 00:10:28.840 --> 00:10:30.840 align:middle line:84% And we'll have a slide coming up soon 00:10:30.840 --> 00:10:34.470 align:middle line:84% of how this R value is calculated. 00:10:34.470 --> 00:10:35.100 align:middle line:90% OK. 00:10:35.100 --> 00:10:37.740 align:middle line:84% The next line down begins the sequence. 00:10:37.740 --> 00:10:41.520 align:middle line:84% And if you have a multi-chain sequence, here I've cut-- 00:10:41.520 --> 00:10:43.290 align:middle line:90% I've cut out some lines for-- 00:10:43.290 --> 00:10:45.630 align:middle line:84% there's many lines of sequence for the protein 00:10:45.630 --> 00:10:47.520 align:middle line:84% in this three-letter code, and main lines 00:10:47.520 --> 00:10:50.310 align:middle line:84% of sequence for the nucleic acid in this one-letter code. 00:10:50.310 --> 00:10:54.090 align:middle line:84% And then, additional chemical parts of the structure. 00:10:54.090 --> 00:10:55.960 align:middle line:84% Remember, the structure is complicated. 00:10:55.960 --> 00:10:57.360 align:middle line:90% It's not just protein. 00:10:57.360 --> 00:10:59.850 align:middle line:84% Here it has nucleic acid, has zinc. 00:10:59.850 --> 00:11:03.150 align:middle line:84% It has water molecules, sometimes various other things. 00:11:03.150 --> 00:11:07.260 align:middle line:84% Each of these molecules, if you can find it in the structure, 00:11:07.260 --> 00:11:09.840 align:middle line:84% you will determine the X, Y and Z-coordinates. 00:11:09.840 --> 00:11:12.360 align:middle line:84% So the next one tells you the secondary structure. 00:11:12.360 --> 00:11:15.330 align:middle line:84% Remember, that there's three basic types-- alpha helices, 00:11:15.330 --> 00:11:17.370 align:middle line:90% beta sheets, and coils. 00:11:17.370 --> 00:11:20.460 align:middle line:84% And these are described-- and again, I'm just showing you 00:11:20.460 --> 00:11:22.180 align:middle line:90% one line example of each. 00:11:22.180 --> 00:11:25.110 align:middle line:84% There's a long list for each of these, where they've 00:11:25.110 --> 00:11:27.390 align:middle line:84% been identified by either manually 00:11:27.390 --> 00:11:30.930 align:middle line:84% or a computational automation from the structure. 00:11:30.930 --> 00:11:37.320 align:middle line:84% And these can be useful as a summary of the structure. 00:11:37.320 --> 00:11:39.960 align:middle line:84% And then, here's the real meat of the structure. 00:11:39.960 --> 00:11:43.140 align:middle line:84% The lines that begin with the word atom, this 00:11:43.140 --> 00:11:45.900 align:middle line:84% is a position of the nitrogen atom number 00:11:45.900 --> 00:11:49.110 align:middle line:84% one in methionine, which is the amino acid number 00:11:49.110 --> 00:11:51.570 align:middle line:90% one in the A chain. 00:11:51.570 --> 00:11:57.990 align:middle line:84% So it met A. The A-chain happens to be the protein chain. 00:11:57.990 --> 00:12:03.221 align:middle line:84% And then, following that is the residue number one. 00:12:03.221 --> 00:12:03.721 align:middle line:90% 00:12:03.721 --> 00:12:08.760 align:middle line:84% And then, XYZ coordinates, roughly 50, 24, 79. 00:12:08.760 --> 00:12:11.880 align:middle line:84% Then a scale factor one, which is almost always one. 00:12:11.880 --> 00:12:14.700 align:middle line:84% And then a B factor, 60, which is 00:12:14.700 --> 00:12:20.220 align:middle line:84% representative of how far from that XYZ value can it deviate? 00:12:20.220 --> 00:12:24.330 align:middle line:90% That's a square deviation term. 00:12:24.330 --> 00:12:29.610 align:middle line:84% And it absorbs the thermal motion 00:12:29.610 --> 00:12:32.620 align:middle line:84% of that atom and various structural defects. 00:12:32.620 --> 00:12:35.970 align:middle line:84% So it gives you some idea of the disorder of that atom. 00:12:35.970 --> 00:12:39.540 align:middle line:84% And then you have the last couple of records have to do, 00:12:39.540 --> 00:12:44.760 align:middle line:84% what atom is connected to what atom in the structure? 00:12:44.760 --> 00:12:46.260 align:middle line:84% In a certain sense, those can often 00:12:46.260 --> 00:12:50.570 align:middle line:84% be inferred just by the distance between atoms in the structure. 00:12:50.570 --> 00:12:52.650 align:middle line:84% Now on the far right hand side, it's 00:12:52.650 --> 00:12:58.620 align:middle line:84% just the record number and the shorthand for the structure, 00:12:58.620 --> 00:13:00.330 align:middle line:90% which is 1/8 CQ. 00:13:00.330 --> 00:13:04.620 align:middle line:84% 1/8 CQ refers to the human chicken receptor. 00:13:04.620 --> 00:13:07.800 align:middle line:90% 00:13:07.800 --> 00:13:09.840 align:middle line:90% So that's a very dry-- 00:13:09.840 --> 00:13:14.160 align:middle line:84% that's the way that it appears when you download it 00:13:14.160 --> 00:13:19.950 align:middle line:90% from the database PDB or RCSV. 00:13:19.950 --> 00:13:23.790 align:middle line:84% Then when you display it, while you're solving it, 00:13:23.790 --> 00:13:26.430 align:middle line:84% if you're NMR or X-ray crystallographer 00:13:26.430 --> 00:13:29.970 align:middle line:84% or possibly display it from the databases, 00:13:29.970 --> 00:13:33.720 align:middle line:84% and the two different cultures in [INAUDIBLE] on the left 00:13:33.720 --> 00:13:37.800 align:middle line:84% tends to describe their structures as multiple chain 00:13:37.800 --> 00:13:41.340 align:middle line:84% tracings because they want to either express 00:13:41.340 --> 00:13:43.800 align:middle line:84% their uncertainty of the structure, 00:13:43.800 --> 00:13:46.560 align:middle line:84% or they want to brag about how they know 00:13:46.560 --> 00:13:48.030 align:middle line:90% something about the dynamics. 00:13:48.030 --> 00:13:50.610 align:middle line:84% Whatever, you have multiple chains 00:13:50.610 --> 00:13:53.640 align:middle line:84% which overlap here in different colors indicating 00:13:53.640 --> 00:13:55.350 align:middle line:84% some of the different uncertainty 00:13:55.350 --> 00:13:59.130 align:middle line:90% or dynamics of each major atom. 00:13:59.130 --> 00:14:01.680 align:middle line:84% Sometimes you'll show all the atoms as one on the right, 00:14:01.680 --> 00:14:03.330 align:middle line:84% or you'll just show the major atoms 00:14:03.330 --> 00:14:06.660 align:middle line:84% like the carbon alphas which is the center of each amino acid 00:14:06.660 --> 00:14:08.880 align:middle line:90% as you go along on the left. 00:14:08.880 --> 00:14:12.030 align:middle line:84% On the right is the way that X-ray crystallographer 00:14:12.030 --> 00:14:15.910 align:middle line:90% might show the fit to the data. 00:14:15.910 --> 00:14:19.680 align:middle line:84% So the model is a stick figure connecting the atoms 00:14:19.680 --> 00:14:21.660 align:middle line:90% as circles. 00:14:21.660 --> 00:14:24.580 align:middle line:84% And then the mesh work is the electron density, 00:14:24.580 --> 00:14:29.010 align:middle line:84% which you can observe once you have all the X-ray 00:14:29.010 --> 00:14:33.940 align:middle line:84% data and the model, or you can calculate it 00:14:33.940 --> 00:14:35.010 align:middle line:90% once you have the model. 00:14:35.010 --> 00:14:39.372 align:middle line:84% The model plus the known physics of scattering of each of the-- 00:14:39.372 --> 00:14:41.830 align:middle line:84% known physics of the electron density of each of the atoms, 00:14:41.830 --> 00:14:43.980 align:middle line:84% you can calculate electron density. 00:14:43.980 --> 00:14:47.610 align:middle line:84% Now you can compare the calculated electron density 00:14:47.610 --> 00:14:48.480 align:middle line:90% with the observed. 00:14:48.480 --> 00:14:51.390 align:middle line:84% Or you can compare the calculated scattering 00:14:51.390 --> 00:14:53.520 align:middle line:90% with the observed. 00:14:53.520 --> 00:14:55.570 align:middle line:84% Typically, it's done in the scattering, 00:14:55.570 --> 00:14:58.440 align:middle line:84% which is the 4E transform of the electron density. 00:14:58.440 --> 00:15:00.630 align:middle line:90% And that's all this is. 00:15:00.630 --> 00:15:03.090 align:middle line:84% The electron density is indicated by RHO here 00:15:03.090 --> 00:15:04.770 align:middle line:90% in the middle of this formula. 00:15:04.770 --> 00:15:07.800 align:middle line:84% And the 4E transform is just this integral 00:15:07.800 --> 00:15:12.000 align:middle line:84% of RHO over the phasing information, 00:15:12.000 --> 00:15:16.680 align:middle line:90% the phasing of the light waves. 00:15:16.680 --> 00:15:20.010 align:middle line:84% These waves, just like a wave on an ocean has a phase. 00:15:20.010 --> 00:15:22.870 align:middle line:84% Whether it's up or down in the trough, how much. 00:15:22.870 --> 00:15:27.040 align:middle line:84% And so the product of the RHO electron density, 00:15:27.040 --> 00:15:30.180 align:middle line:84% which is a function of X, Y, and Z, all three coordinates, 00:15:30.180 --> 00:15:33.060 align:middle line:84% its is summed by these integrals-- 00:15:33.060 --> 00:15:37.310 align:middle line:84% It's a continuous function-- from 0 to 1. 00:15:37.310 --> 00:15:40.268 align:middle line:84% 0 to 1, and X, Y, and Z. This is Y01, 00:15:40.268 --> 00:15:42.060 align:middle line:84% it's because this is a repeating structure. 00:15:42.060 --> 00:15:48.660 align:middle line:84% It has a little rectangular cube of space around it, 00:15:48.660 --> 00:15:49.595 align:middle line:90% which repeats. 00:15:49.595 --> 00:15:51.470 align:middle line:84% And so all you really need to do to calculate 00:15:51.470 --> 00:15:53.480 align:middle line:84% the entire electron density is to think 00:15:53.480 --> 00:15:55.720 align:middle line:84% about this little cube, which goes from 0 00:15:55.720 --> 00:15:57.140 align:middle line:90% to 1 in those arbitrary units. 00:15:57.140 --> 00:15:59.690 align:middle line:90% 00:15:59.690 --> 00:16:02.930 align:middle line:84% But now that's how you can get from one space to another. 00:16:02.930 --> 00:16:05.030 align:middle line:84% From the electron density to the scattering 00:16:05.030 --> 00:16:07.160 align:middle line:84% that you actually observe when you 00:16:07.160 --> 00:16:11.760 align:middle line:84% shine x-ray light upon a repeating crystal structure. 00:16:11.760 --> 00:16:13.460 align:middle line:84% But now you want to adjust the model. 00:16:13.460 --> 00:16:16.740 align:middle line:84% Adjust those atoms in the previous slide 00:16:16.740 --> 00:16:20.400 align:middle line:84% so that you can maximize the fit. 00:16:20.400 --> 00:16:22.170 align:middle line:84% That is to say, minimize the difference 00:16:22.170 --> 00:16:24.930 align:middle line:84% between the observed scattering F0 00:16:24.930 --> 00:16:26.753 align:middle line:84% and the calculated scattering, FC. 00:16:26.753 --> 00:16:28.920 align:middle line:84% Because you know the scattering of each atom and you 00:16:28.920 --> 00:16:30.750 align:middle line:84% know that you're trying to determine 00:16:30.750 --> 00:16:32.070 align:middle line:90% the position of each atom. 00:16:32.070 --> 00:16:36.870 align:middle line:84% The position of the atom might be a parameter, P, 00:16:36.870 --> 00:16:39.390 align:middle line:84% which you adjust a small bit at a time. 00:16:39.390 --> 00:16:41.990 align:middle line:90% 00:16:41.990 --> 00:16:46.730 align:middle line:84% And that and that change can be approximated 00:16:46.730 --> 00:16:50.060 align:middle line:84% by what's called a Paler Series Expansion. 00:16:50.060 --> 00:16:52.010 align:middle line:84% Here, we're just taking the first term, which 00:16:52.010 --> 00:16:54.590 align:middle line:84% involves first derivatives, all sub-sequence terms 00:16:54.590 --> 00:16:56.270 align:middle line:84% involve second derivatives and higher. 00:16:56.270 --> 00:17:00.470 align:middle line:84% And those are close enough to 0 for this work 00:17:00.470 --> 00:17:02.750 align:middle line:90% that you drop them. 00:17:02.750 --> 00:17:04.730 align:middle line:84% And this basically says that if we're 00:17:04.730 --> 00:17:06.890 align:middle line:84% going to adjust this parameter, we 00:17:06.890 --> 00:17:10.560 align:middle line:84% can get a feeling for how fast to adjust it, 00:17:10.560 --> 00:17:13.910 align:middle line:84% which is based on the sensitivity of the scattering, 00:17:13.910 --> 00:17:16.940 align:middle line:84% the F. The derivative of F calculated 00:17:16.940 --> 00:17:18.859 align:middle line:90% with respect to each parameter. 00:17:18.859 --> 00:17:21.230 align:middle line:84% The parameters would be X, Y, and Z-coordinates, 00:17:21.230 --> 00:17:25.972 align:middle line:84% or they could be some kind of rotational parameters. 00:17:25.972 --> 00:17:28.430 align:middle line:84% So this is to give you a flavor for how it's actually done. 00:17:28.430 --> 00:17:32.030 align:middle line:84% This is how you actually get from the scattering off 00:17:32.030 --> 00:17:32.630 align:middle line:90% of a crystal. 00:17:32.630 --> 00:17:33.980 align:middle line:90% The crystal has the advantage. 00:17:33.980 --> 00:17:35.980 align:middle line:84% In principle, you can do a scattering experiment 00:17:35.980 --> 00:17:37.490 align:middle line:90% from single molecules. 00:17:37.490 --> 00:17:40.280 align:middle line:84% But single molecules, the signal is too weak. 00:17:40.280 --> 00:17:44.630 align:middle line:84% And it's swamped out by the noise of random other photon 00:17:44.630 --> 00:17:46.040 align:middle line:90% events. 00:17:46.040 --> 00:17:49.460 align:middle line:84% So by having a large number of them in ordered array, 00:17:49.460 --> 00:17:52.880 align:middle line:84% they all cohere and they basically 00:17:52.880 --> 00:17:54.350 align:middle line:90% do your statistics for you. 00:17:54.350 --> 00:17:59.510 align:middle line:84% They integrate and you get the value 00:17:59.510 --> 00:18:02.032 align:middle line:84% of the statistics of billions of molecules 00:18:02.032 --> 00:18:04.490 align:middle line:84% without having to observe each of the billions of molecules 00:18:04.490 --> 00:18:07.550 align:middle line:84% and then do the computation in the presence of a huge noise. 00:18:07.550 --> 00:18:09.800 align:middle line:84% So that's what the crystal is all about. 00:18:09.800 --> 00:18:12.150 align:middle line:84% NMR also requires billions of molecules. 00:18:12.150 --> 00:18:13.760 align:middle line:84% And so they both have the big demand 00:18:13.760 --> 00:18:18.470 align:middle line:84% of requiring large amounts of pure molecules. 00:18:18.470 --> 00:18:20.660 align:middle line:84% And that's one of the reasons that membrane proteins 00:18:20.660 --> 00:18:22.550 align:middle line:90% have been harder to get at. 00:18:22.550 --> 00:18:26.810 align:middle line:84% It's harder to get large amounts of pure membrane pieces. 00:18:26.810 --> 00:18:30.650 align:middle line:84% Now these two methods NMR, which I 00:18:30.650 --> 00:18:33.380 align:middle line:84% won't describe, and extra crystallography that I barely 00:18:33.380 --> 00:18:36.530 align:middle line:84% described, share with the abinitio methods 00:18:36.530 --> 00:18:42.650 align:middle line:84% of protein structure prediction certain key computational 00:18:42.650 --> 00:18:44.790 align:middle line:90% components. 00:18:44.790 --> 00:18:47.930 align:middle line:84% And these are embedded in a combined system, which 00:18:47.930 --> 00:18:50.060 align:middle line:84% does crystallography, NMR, and some 00:18:50.060 --> 00:18:52.340 align:middle line:84% of these molecular mechanics that keeps the structure. 00:18:52.340 --> 00:18:56.930 align:middle line:84% You can imagine that if the structure started blowing off, 00:18:56.930 --> 00:18:58.550 align:middle line:84% atoms going in weird directions, it 00:18:58.550 --> 00:19:00.660 align:middle line:84% could still minimize the function 00:19:00.660 --> 00:19:02.040 align:middle line:90% if you have a local minimum. 00:19:02.040 --> 00:19:04.910 align:middle line:84% But if you hold the chemistry intact 00:19:04.910 --> 00:19:07.340 align:middle line:84% and satisfy what you know about molecule mechanics 00:19:07.340 --> 00:19:11.030 align:middle line:84% of chemicals, then you actually can fit a structure 00:19:11.030 --> 00:19:13.700 align:middle line:90% from further away. 00:19:13.700 --> 00:19:16.280 align:middle line:84% Now here's the ARC factor I said we will come back to. 00:19:16.280 --> 00:19:17.960 align:middle line:84% As you do this refinement, as you 00:19:17.960 --> 00:19:20.090 align:middle line:84% adjust the positions of each of the atoms, 00:19:20.090 --> 00:19:22.190 align:middle line:84% as your computer adjusts the position of each 00:19:22.190 --> 00:19:25.910 align:middle line:84% of these atoms, you compare the scattering 00:19:25.910 --> 00:19:30.680 align:middle line:90% of the observed F0 with the FC. 00:19:30.680 --> 00:19:32.750 align:middle line:84% These are in absolute values because actually, 00:19:32.750 --> 00:19:36.110 align:middle line:84% the scattering results in loss of phase information. 00:19:36.110 --> 00:19:39.140 align:middle line:84% So the actual things you measure are absolute values. 00:19:39.140 --> 00:19:41.390 align:middle line:84% And then you take the absolute value of the difference 00:19:41.390 --> 00:19:45.980 align:middle line:84% in order to make sure you sum a positive number. 00:19:45.980 --> 00:19:50.360 align:middle line:84% And then you normalize this, as we did before, 00:19:50.360 --> 00:19:55.375 align:middle line:84% to put it on a recognizable standard scale where 0.4 means 00:19:55.375 --> 00:19:56.750 align:middle line:84% you have a very crude structure-- 00:19:56.750 --> 00:19:58.370 align:middle line:84% if you see this in the literature you don't believe 00:19:58.370 --> 00:19:58.870 align:middle line:90% it. 00:19:58.870 --> 00:20:02.510 align:middle line:84% If it's less than 0.25, which the last structure was, 00:20:02.510 --> 00:20:06.620 align:middle line:84% then you believe it as pretty close to done. 00:20:06.620 --> 00:20:08.808 align:middle line:84% This is very analogous to correlation coefficient. 00:20:08.808 --> 00:20:10.850 align:middle line:84% Remember, we had a linear correlation coefficient 00:20:10.850 --> 00:20:12.110 align:middle line:84% between two functions, here would 00:20:12.110 --> 00:20:13.280 align:middle line:90% be observed and calculated. 00:20:13.280 --> 00:20:16.730 align:middle line:84% If they correlate well, then you're getting close to done. 00:20:16.730 --> 00:20:20.180 align:middle line:84% Correlating well is better than 0.7, in this case. 00:20:20.180 --> 00:20:23.713 align:middle line:84% And one way of reporting the similarities between two 00:20:23.713 --> 00:20:25.130 align:middle line:84% structures, this is not a goodness 00:20:25.130 --> 00:20:26.588 align:middle line:84% of fit between model and data, this 00:20:26.588 --> 00:20:29.390 align:middle line:84% is a goodness of fit between two models. 00:20:29.390 --> 00:20:32.300 align:middle line:84% Atom by atom you go through and you measure 00:20:32.300 --> 00:20:34.910 align:middle line:90% the distance between them. 00:20:34.910 --> 00:20:38.290 align:middle line:84% And that root mean squared deviation, 00:20:38.290 --> 00:20:40.300 align:middle line:84% of all the distances over all the atoms, 00:20:40.300 --> 00:20:44.110 align:middle line:84% or all the key atoms, core atoms, carbon alphas-- 00:20:44.110 --> 00:20:46.870 align:middle line:84% or maybe even a smaller core than that-- 00:20:46.870 --> 00:20:51.940 align:middle line:84% allows you to quote a root mean square deviation, which 00:20:51.940 --> 00:20:56.470 align:middle line:84% has some meaning independent of how many atoms you have 00:20:56.470 --> 00:20:58.210 align:middle line:84% and what proteins you're looking at. 00:20:58.210 --> 00:21:00.460 align:middle line:84% Each of these is try to put this on a common scale 00:21:00.460 --> 00:21:04.905 align:middle line:84% so you can compare from structure to structure. 00:21:04.905 --> 00:21:07.030 align:middle line:84% Now if we're going to do molecular mechanics, which 00:21:07.030 --> 00:21:12.670 align:middle line:84% is common to the computational empirical methods 00:21:12.670 --> 00:21:15.490 align:middle line:84% and the computational sequence based methods, 00:21:15.490 --> 00:21:23.620 align:middle line:84% we need to talk about the side chains of the proteins. 00:21:23.620 --> 00:21:26.590 align:middle line:84% We've mainly been talking about the backbones. 00:21:26.590 --> 00:21:28.990 align:middle line:84% And just a refresher, this is from the geniculate code. 00:21:28.990 --> 00:21:31.150 align:middle line:84% Again, the blue or the positively charge 00:21:31.150 --> 00:21:32.860 align:middle line:84% and the negative charge and so forth. 00:21:32.860 --> 00:21:35.290 align:middle line:90% They have a chirality. 00:21:35.290 --> 00:21:38.170 align:middle line:84% It matters whether you're talking about L-amino acids 00:21:38.170 --> 00:21:39.250 align:middle line:90% or D-amino acids. 00:21:39.250 --> 00:21:40.840 align:middle line:84% The way you remember-- this is just 00:21:40.840 --> 00:21:42.670 align:middle line:84% a mnemonic for remembering it-- is 00:21:42.670 --> 00:21:45.250 align:middle line:84% that when the hydrogens point towards you, going clockwise, 00:21:45.250 --> 00:21:51.340 align:middle line:84% it goes C0 carbonyl R, this is a side chain in corn. 00:21:51.340 --> 00:21:56.140 align:middle line:84% And some of the 19 of these amino acids 00:21:56.140 --> 00:22:00.640 align:middle line:90% have a chirality there. 00:22:00.640 --> 00:22:04.870 align:middle line:84% Glycine does not because it has two hydrogens instead of an R, 00:22:04.870 --> 00:22:08.140 align:middle line:84% it has a hydrogen. And two of the amino acids 00:22:08.140 --> 00:22:11.020 align:middle line:84% actually have two centers of chiral asymmetry. 00:22:11.020 --> 00:22:19.150 align:middle line:84% Threonine, which has this side chain, and Isoleucine. 00:22:19.150 --> 00:22:21.310 align:middle line:84% They have the carbon alpha and the carbon beta 00:22:21.310 --> 00:22:23.290 align:middle line:90% are both asymmetric. 00:22:23.290 --> 00:22:25.480 align:middle line:84% And one of the very earliest exercises 00:22:25.480 --> 00:22:28.720 align:middle line:84% was done when the very first models of proteins 00:22:28.720 --> 00:22:30.540 align:middle line:90% were looked at, little peptides. 00:22:30.540 --> 00:22:33.880 align:middle line:90% 00:22:33.880 --> 00:22:37.690 align:middle line:84% Can do this by hand, with some very simple crude models. 00:22:37.690 --> 00:22:40.210 align:middle line:84% And you can go through systematically, 00:22:40.210 --> 00:22:45.820 align:middle line:84% there are three bonds along a peptide, long peptide. 00:22:45.820 --> 00:22:48.230 align:middle line:84% And these are the peptide bond itself, 00:22:48.230 --> 00:22:50.980 align:middle line:84% which connects one amino acid to the next one. 00:22:50.980 --> 00:22:52.440 align:middle line:84% And that tends to be pretty rigid. 00:22:52.440 --> 00:22:54.410 align:middle line:84% It has a partial double stranded bond 00:22:54.410 --> 00:22:57.010 align:middle line:84% and it tends to be a trans-configuration, 180 00:22:57.010 --> 00:22:57.920 align:middle line:90% degrees. 00:22:57.920 --> 00:23:00.770 align:middle line:84% This is the rotation around the bonds, not the bond angle, 00:23:00.770 --> 00:23:02.470 align:middle line:84% but the rotation around the bonds. 00:23:02.470 --> 00:23:07.000 align:middle line:84% And there are two other bonds that are not so rigid. 00:23:07.000 --> 00:23:09.490 align:middle line:84% So these are free, but they're constrained 00:23:09.490 --> 00:23:12.250 align:middle line:84% by the clashes that occur that when you rotate around 00:23:12.250 --> 00:23:16.960 align:middle line:84% the bond, the side chains will clash 00:23:16.960 --> 00:23:19.220 align:middle line:90% with other parts of the protein. 00:23:19.220 --> 00:23:21.850 align:middle line:84% And so Ramachandran and colleagues went through 00:23:21.850 --> 00:23:25.100 align:middle line:84% systematically, all the possible Phi and PSI angles, 00:23:25.100 --> 00:23:26.950 align:middle line:90% these are these two free bonds. 00:23:26.950 --> 00:23:30.850 align:middle line:84% And this is shown here ranging over the full range of Phi 00:23:30.850 --> 00:23:33.850 align:middle line:84% and PSI on the horizontal vertical axis. 00:23:33.850 --> 00:23:35.620 align:middle line:84% And you get these little orange regions 00:23:35.620 --> 00:23:40.120 align:middle line:84% where even with very bulky chain groups, which would occur 00:23:40.120 --> 00:23:42.190 align:middle line:90% you get these allowed regions. 00:23:42.190 --> 00:23:44.290 align:middle line:84% And these two allowed regions happened 00:23:44.290 --> 00:23:47.500 align:middle line:84% to coincide with two of the most popular motifs you 00:23:47.500 --> 00:23:50.920 align:middle line:84% find in proteins, which are the beta sheet and the alpha helix. 00:23:50.920 --> 00:23:55.480 align:middle line:90% 00:23:55.480 --> 00:23:57.610 align:middle line:84% There are other things, such as the 310 helix 00:23:57.610 --> 00:23:59.500 align:middle line:84% and various other structures that turn up. 00:23:59.500 --> 00:24:01.720 align:middle line:84% But those are by far the two most common. 00:24:01.720 --> 00:24:03.130 align:middle line:84% And the yellow shows how they get 00:24:03.130 --> 00:24:05.088 align:middle line:84% extended when you have smaller side chains that 00:24:05.088 --> 00:24:09.580 align:middle line:84% allow more parts of the conformation spaces 00:24:09.580 --> 00:24:12.430 align:middle line:90% as it's called, to be inspected. 00:24:12.430 --> 00:24:14.320 align:middle line:84% Now that's a very crude thing that you can do 00:24:14.320 --> 00:24:16.105 align:middle line:90% with very simple stick figures. 00:24:16.105 --> 00:24:20.340 align:middle line:90% 00:24:20.340 --> 00:24:24.360 align:middle line:84% But as you get to more detailed analysis, 00:24:24.360 --> 00:24:28.170 align:middle line:84% the ultimate application of all we know about physics, 00:24:28.170 --> 00:24:32.040 align:middle line:84% if we could compute them would be quantum electrodynamics. 00:24:32.040 --> 00:24:35.730 align:middle line:84% This is way out of range for any molecule of the size 00:24:35.730 --> 00:24:37.150 align:middle line:90% that we're interested in. 00:24:37.150 --> 00:24:38.730 align:middle line:84% And then as you go down this list, 00:24:38.730 --> 00:24:40.590 align:middle line:84% you get more and more precise programs 00:24:40.590 --> 00:24:42.600 align:middle line:84% until you get down to something which 00:24:42.600 --> 00:24:45.720 align:middle line:84% is barely computationally feasible for things 00:24:45.720 --> 00:24:46.710 align:middle line:90% the size of proteins. 00:24:46.710 --> 00:24:52.050 align:middle line:84% And a great approximation of all the quantum 00:24:52.050 --> 00:24:53.275 align:middle line:90% approximations above it. 00:24:53.275 --> 00:24:54.900 align:middle line:84% Every one of these is an approximation, 00:24:54.900 --> 00:24:58.115 align:middle line:84% but each one as you go down, gets more and more approximate. 00:24:58.115 --> 00:24:59.490 align:middle line:84% And the main thing that's missing 00:24:59.490 --> 00:25:02.130 align:middle line:84% from molecular mechanics that's present in the next step 00:25:02.130 --> 00:25:04.920 align:middle line:84% up is the polarization of electrons. 00:25:04.920 --> 00:25:07.770 align:middle line:84% In other words, in micro mechanics, 00:25:07.770 --> 00:25:10.740 align:middle line:84% you assume the electron clouds are basically spherical. 00:25:10.740 --> 00:25:16.830 align:middle line:84% And this is a huge loss, but it still is computationally very 00:25:16.830 --> 00:25:17.580 align:middle line:90% demanding. 00:25:17.580 --> 00:25:20.310 align:middle line:84% So you don't get that asymmetric polarization 00:25:20.310 --> 00:25:27.880 align:middle line:84% that you get in hydrogen bonds and many other dipoles. 00:25:27.880 --> 00:25:30.690 align:middle line:90% So this is really-- 00:25:30.690 --> 00:25:32.280 align:middle line:90% this is basic physics. 00:25:32.280 --> 00:25:34.560 align:middle line:84% You can see the first line, forced equals mass times 00:25:34.560 --> 00:25:35.340 align:middle line:90% acceleration. 00:25:35.340 --> 00:25:38.250 align:middle line:90% Basic Newton's law. 00:25:38.250 --> 00:25:42.090 align:middle line:84% And Newton also introduced the calculus to us. 00:25:42.090 --> 00:25:47.490 align:middle line:84% So he would be very comfortable with the next line, which 00:25:47.490 --> 00:25:51.870 align:middle line:84% is that force can be redefined as the first derivative 00:25:51.870 --> 00:25:57.420 align:middle line:84% of energy with respect to position or radius. 00:25:57.420 --> 00:25:59.700 align:middle line:90% And then mass is just mass. 00:25:59.700 --> 00:26:02.280 align:middle line:84% And we're introducing the subscript I, for the atomic. 00:26:02.280 --> 00:26:06.780 align:middle line:84% For each atom gets its own Newton's law. 00:26:06.780 --> 00:26:11.070 align:middle line:84% And then acceleration is just a second derivative of position 00:26:11.070 --> 00:26:13.870 align:middle line:90% with respect to time. 00:26:13.870 --> 00:26:16.390 align:middle line:84% Now what kind of time constants are we talking about here? 00:26:16.390 --> 00:26:19.530 align:middle line:84% This is the femtosecond range for atomic motion 00:26:19.530 --> 00:26:21.100 align:middle line:90% to the minus 15 seconds. 00:26:21.100 --> 00:26:25.710 align:middle line:84% And as you step through, you update this kinetic procedure, 00:26:25.710 --> 00:26:27.990 align:middle line:84% you can do it in half time intervals, 00:26:27.990 --> 00:26:33.780 align:middle line:84% updating velocity and position every femtosecond 00:26:33.780 --> 00:26:35.940 align:middle line:90% or half femtosecond. 00:26:35.940 --> 00:26:37.710 align:middle line:90% So now what's this energy term? 00:26:37.710 --> 00:26:40.470 align:middle line:84% This is what I alluded to in the previous slide 00:26:40.470 --> 00:26:42.060 align:middle line:90% as being very approximate. 00:26:42.060 --> 00:26:44.760 align:middle line:84% And semi-empirical, it is based on experiments 00:26:44.760 --> 00:26:46.620 align:middle line:84% not entirely from first principles 00:26:46.620 --> 00:26:49.170 align:middle line:84% or not even from the quantum approximations. 00:26:49.170 --> 00:26:52.620 align:middle line:84% You have, say, spectroscopic analyzes 00:26:52.620 --> 00:26:58.110 align:middle line:84% that will show that the spring-like motion that two 00:26:58.110 --> 00:27:00.810 align:middle line:84% atoms can have when they're connected by a bond 00:27:00.810 --> 00:27:05.500 align:middle line:84% has a kind of a Hooke's law type of spring motion. 00:27:05.500 --> 00:27:09.420 align:middle line:84% And that's the energy of the bond length, 00:27:09.420 --> 00:27:13.980 align:middle line:84% EB in this sum of all the E's, slide 52. 00:27:13.980 --> 00:27:19.680 align:middle line:84% And E theta is the angle that you have as a bond angle bends. 00:27:19.680 --> 00:27:21.840 align:middle line:84% And that's the spring-like force. 00:27:21.840 --> 00:27:24.437 align:middle line:84% And then omega is this kind of torsion angle 00:27:24.437 --> 00:27:26.520 align:middle line:84% that we've been talking about in the phi PSI plot, 00:27:26.520 --> 00:27:28.800 align:middle line:90% Ramachandran plot just before. 00:27:28.800 --> 00:27:30.990 align:middle line:84% Van der Waals is the non-bonded contact, 00:27:30.990 --> 00:27:32.710 align:middle line:84% which can be either positive or negative. 00:27:32.710 --> 00:27:34.710 align:middle line:84% Actually, it should show down at the very bottom 00:27:34.710 --> 00:27:39.480 align:middle line:84% of the slide is that there is a repulsive force, which 00:27:39.480 --> 00:27:44.130 align:middle line:84% is related to the R to the 12th power. 00:27:44.130 --> 00:27:47.530 align:middle line:84% And an attractive force, which is R to the sixth power. 00:27:47.530 --> 00:27:49.860 align:middle line:84% So as you get closer, it starts to get attracted 00:27:49.860 --> 00:27:53.010 align:middle line:84% until you get this hard sphere repulsion as you 00:27:53.010 --> 00:27:54.150 align:middle line:90% get a little bit closer. 00:27:54.150 --> 00:27:58.230 align:middle line:84% Electrostatic interactions are the longest range effects. 00:27:58.230 --> 00:28:03.030 align:middle line:84% All these covalent bonds, B and theta and omega, 00:28:03.030 --> 00:28:03.910 align:middle line:90% are short range. 00:28:03.910 --> 00:28:05.160 align:middle line:90% Van der Waals are short range. 00:28:05.160 --> 00:28:09.180 align:middle line:84% Electrostatic is slightly longer range because it's a 1 over R, 00:28:09.180 --> 00:28:12.190 align:middle line:84% or R is the distance between the two atoms. 00:28:12.190 --> 00:28:13.920 align:middle line:84% And those are the main terms that 00:28:13.920 --> 00:28:15.510 align:middle line:84% enter into all molecular mechanics, 00:28:15.510 --> 00:28:17.135 align:middle line:84% whether they're used in crystallography 00:28:17.135 --> 00:28:18.900 align:middle line:84% or whether they used in abinitio. 00:28:18.900 --> 00:28:21.480 align:middle line:84% Now this is the state of the art for abinitio. 00:28:21.480 --> 00:28:26.070 align:middle line:84% Just the very most recent CAS competition 00:28:26.070 --> 00:28:30.660 align:middle line:84% resulted in a very clear winner by some criteria, at least. 00:28:30.660 --> 00:28:34.500 align:middle line:84% The Baker Lab here, the URL is down here, 00:28:34.500 --> 00:28:36.150 align:middle line:84% is the number of standard deviations 00:28:36.150 --> 00:28:40.650 align:middle line:84% away for the mean in terms of the score for the number 00:28:40.650 --> 00:28:43.290 align:middle line:84% of correct predictions, here out at around 30 00:28:43.290 --> 00:28:46.830 align:middle line:90% where the mean is close to zero. 00:28:46.830 --> 00:28:53.400 align:middle line:84% And even with this huge advance for the field in prediction, 00:28:53.400 --> 00:28:57.690 align:middle line:84% still this is a typical RMS standard deviation between 00:28:57.690 --> 00:29:00.270 align:middle line:84% the real structure, which was kept hidden from sight from all 00:29:00.270 --> 00:29:02.610 align:middle line:84% these competitors-- until it was known, 00:29:02.610 --> 00:29:05.320 align:middle line:90% but not to the competitors-- 00:29:05.320 --> 00:29:06.180 align:middle line:90% and then revealed. 00:29:06.180 --> 00:29:09.750 align:middle line:84% And the RMS deviation was 6, or 4, or 5 00:29:09.750 --> 00:29:13.200 align:middle line:84% in that range depending on the structure 00:29:13.200 --> 00:29:17.310 align:middle line:84% and whether you include all the atoms or just the core ones. 00:29:17.310 --> 00:29:20.070 align:middle line:84% And this is not adequate, as we saw in that slide 00:29:20.070 --> 00:29:23.310 align:middle line:84% actually from the same group earlier on. 00:29:23.310 --> 00:29:26.130 align:middle line:90% 00:29:26.130 --> 00:29:28.230 align:middle line:84% Another way of looking at this is now-- 00:29:28.230 --> 00:29:31.620 align:middle line:84% those were predictive structures-- these are now 00:29:31.620 --> 00:29:32.760 align:middle line:90% observed structures. 00:29:32.760 --> 00:29:37.110 align:middle line:84% The purple is comparing two structures, both of which 00:29:37.110 --> 00:29:39.300 align:middle line:84% were done by X-ray crystallography. 00:29:39.300 --> 00:29:44.790 align:middle line:84% And along the red axis here is sequence identity, ranging 00:29:44.790 --> 00:29:48.370 align:middle line:84% from 0 to close to 100%, say 96 plus percent. 00:29:48.370 --> 00:29:51.960 align:middle line:84% And the green axis is the RMS root 00:29:51.960 --> 00:29:55.440 align:middle line:84% mean square mediation between structure one 00:29:55.440 --> 00:29:57.420 align:middle line:90% and structure two. 00:29:57.420 --> 00:30:01.560 align:middle line:84% And you can see that-- think of this purple curve 00:30:01.560 --> 00:30:04.650 align:middle line:84% as starting in the lower right, where you have 00:30:04.650 --> 00:30:10.440 align:middle line:84% very high sequence identity and less than 1 angstrom written 00:30:10.440 --> 00:30:12.420 align:middle line:90% in deviation. 00:30:12.420 --> 00:30:16.620 align:middle line:90% 00:30:16.620 --> 00:30:19.357 align:middle line:84% That means that when you solve two proteins that 00:30:19.357 --> 00:30:20.940 align:middle line:84% are very similar in sequence, you will 00:30:20.940 --> 00:30:22.530 align:middle line:90% get very similar structures. 00:30:22.530 --> 00:30:23.160 align:middle line:90% That's good. 00:30:23.160 --> 00:30:25.170 align:middle line:84% That bodes well for structural modeling, 00:30:25.170 --> 00:30:26.880 align:middle line:84% although that is not structural modeling, 00:30:26.880 --> 00:30:28.740 align:middle line:90% I mean, not homology modeling. 00:30:28.740 --> 00:30:31.320 align:middle line:84% Then as you go down in sequence identity, 00:30:31.320 --> 00:30:33.600 align:middle line:84% the purple curve starts to slope up and up 00:30:33.600 --> 00:30:39.060 align:middle line:84% until it starts curving up towards 2.74 and beyond. 00:30:39.060 --> 00:30:43.560 align:middle line:84% it gets harder and harder to do these structural alignments. 00:30:43.560 --> 00:30:48.945 align:middle line:84% And so 4 angstroms is the sort you 00:30:48.945 --> 00:30:51.630 align:middle line:84% would get from homology modeling at less than 20% 00:30:51.630 --> 00:30:53.740 align:middle line:90% or 30% sequence identity. 00:30:53.740 --> 00:30:55.380 align:middle line:84% And this is what I said earlier, why 00:30:55.380 --> 00:30:58.080 align:middle line:84% we're trying to get enough proteins populating. 00:30:58.080 --> 00:31:00.510 align:middle line:84% This is all known proteins here being compared, 00:31:00.510 --> 00:31:02.550 align:middle line:90% when enough populating it. 00:31:02.550 --> 00:31:08.820 align:middle line:84% So you never have to go below 35% into this Twilight Zone, 00:31:08.820 --> 00:31:10.830 align:middle line:84% where you really can't make good-- 00:31:10.830 --> 00:31:14.190 align:middle line:90% 00:31:14.190 --> 00:31:19.290 align:middle line:84% you don't find good RMS deviations between two 00:31:19.290 --> 00:31:23.280 align:middle line:90% known crystal structures. 00:31:23.280 --> 00:31:28.980 align:middle line:84% Now as we do protein dynamics using the molecular mechanics 00:31:28.980 --> 00:31:31.620 align:middle line:84% approximation we talked about, these 00:31:31.620 --> 00:31:36.270 align:middle line:84% can be applied not only to predict 00:31:36.270 --> 00:31:42.810 align:middle line:84% a static structure or a series of steps in a protein process, 00:31:42.810 --> 00:31:46.590 align:middle line:84% but the dynamics of folding from a completely unfolded protein 00:31:46.590 --> 00:31:49.110 align:middle line:84% as it might be coming off the ribosome. 00:31:49.110 --> 00:31:51.840 align:middle line:84% And this is something for which there are relatively 00:31:51.840 --> 00:31:53.070 align:middle line:90% few experimental methods. 00:31:53.070 --> 00:31:55.680 align:middle line:84% And so this is clearly a valuable contribution, 00:31:55.680 --> 00:31:59.640 align:middle line:84% but there's a problem with doing a theoretical calculation 00:31:59.640 --> 00:32:03.030 align:middle line:84% that's hard to empirically verify. 00:32:03.030 --> 00:32:07.110 align:middle line:84% But in any case, to do one of the larger tasks 00:32:07.110 --> 00:32:13.650 align:middle line:84% and IBM and others are sinking significant resources 00:32:13.650 --> 00:32:15.600 align:middle line:90% and infrastructure of this. 00:32:15.600 --> 00:32:20.130 align:middle line:84% But doing your femtosecond time scale 00:32:20.130 --> 00:32:23.610 align:middle line:84% over a one microsecond simulation, you can easily 00:32:23.610 --> 00:32:26.310 align:middle line:84% do the math, that's 10 to the minus 6, 00:32:26.310 --> 00:32:29.190 align:middle line:84% divided by 10 to the -15 is about 10 00:32:29.190 --> 00:32:31.050 align:middle line:84% to the ninth such steps, each of which 00:32:31.050 --> 00:32:34.950 align:middle line:84% involves this big calculation that we just went through all 00:32:34.950 --> 00:32:36.750 align:middle line:90% the energy terms log. 00:32:36.750 --> 00:32:38.230 align:middle line:90% But that's been done for this. 00:32:38.230 --> 00:32:40.500 align:middle line:84% And you can see the blue and the red 00:32:40.500 --> 00:32:44.340 align:middle line:84% represent the calculated and the observed structure 00:32:44.340 --> 00:32:46.860 align:middle line:84% at one point in the dynamic simulation. 00:32:46.860 --> 00:32:49.470 align:middle line:90% 00:32:49.470 --> 00:32:52.020 align:middle line:84% When you have a protein three dimensional structure, 00:32:52.020 --> 00:32:54.960 align:middle line:84% you can try to dock it with small molecules. 00:32:54.960 --> 00:32:57.620 align:middle line:84% This could be easier, in principle, 00:32:57.620 --> 00:33:00.120 align:middle line:84% because you can keep both the small molecule and the protein 00:33:00.120 --> 00:33:02.290 align:middle line:84% relatively rigid as you dock them. 00:33:02.290 --> 00:33:05.100 align:middle line:84% There has to be some flexibility hence the name, flex, 00:33:05.100 --> 00:33:07.560 align:middle line:90% for one of these programs. 00:33:07.560 --> 00:33:11.700 align:middle line:84% And overall, the results are intriguing enough 00:33:11.700 --> 00:33:15.060 align:middle line:84% that you might want to use it as an alternative in the few cases 00:33:15.060 --> 00:33:16.560 align:middle line:84% where you have the three dimensional 00:33:16.560 --> 00:33:18.450 align:middle line:84% structure of a protein, but for some reason 00:33:18.450 --> 00:33:20.010 align:middle line:84% you can't solve the three dimensional 00:33:20.010 --> 00:33:22.360 align:middle line:90% structure of the complex. 00:33:22.360 --> 00:33:24.100 align:middle line:84% But you must remember that actually, 00:33:24.100 --> 00:33:27.210 align:middle line:84% even though we cited that the solving of protein structure 00:33:27.210 --> 00:33:30.757 align:middle line:84% might be $100,000, solving a complex once you 00:33:30.757 --> 00:33:32.340 align:middle line:84% have the protein structure is actually 00:33:32.340 --> 00:33:33.507 align:middle line:90% considerably less than that. 00:33:33.507 --> 00:33:36.480 align:middle line:84% But in any case, this is encouraging 00:33:36.480 --> 00:33:41.700 align:middle line:84% where you have in the order of 0.25 to 1.84 as a root mean 00:33:41.700 --> 00:33:44.520 align:middle line:84% squared deviation between the predicted 00:33:44.520 --> 00:33:47.070 align:middle line:84% and the experimental binding modes of small molecule. 00:33:47.070 --> 00:33:51.300 align:middle line:84% You can imagine that to be off by 1.8 angstroms, 00:33:51.300 --> 00:33:53.760 align:middle line:84% it must be docking in roughly the right pocket, 00:33:53.760 --> 00:33:58.290 align:middle line:84% but maybe at the wrong angle or yeah, maybe slightly off. 00:33:58.290 --> 00:34:01.820 align:middle line:90% 00:34:01.820 --> 00:34:06.560 align:middle line:84% So the last topic is the issue of cross talk. 00:34:06.560 --> 00:34:09.170 align:middle line:84% As we talk about protein three dimensional structures, 00:34:09.170 --> 00:34:10.940 align:middle line:90% we try to find homologs. 00:34:10.940 --> 00:34:14.900 align:middle line:84% And we often find homologs within an organism, pair logs. 00:34:14.900 --> 00:34:19.100 align:middle line:84% And these pair logs and alternative splice forms 00:34:19.100 --> 00:34:22.340 align:middle line:84% of a protein are potential toxic side 00:34:22.340 --> 00:34:24.800 align:middle line:90% reactions of a particular drug. 00:34:24.800 --> 00:34:26.630 align:middle line:84% And you can see that many of these drugs 00:34:26.630 --> 00:34:28.880 align:middle line:90% are aimed at family members. 00:34:28.880 --> 00:34:33.320 align:middle line:84% For example, a top two are part of the steroid binding 00:34:33.320 --> 00:34:38.389 align:middle line:84% family, which we have already introduced once 00:34:38.389 --> 00:34:41.909 align:middle line:84% and it will be in an upcoming slide. 00:34:41.909 --> 00:34:51.338 align:middle line:84% And when you consider that these proteins, that particular class 00:34:51.338 --> 00:34:53.630 align:middle line:84% of proteins interacts both with a small molecule, which 00:34:53.630 --> 00:34:58.280 align:middle line:84% is either a natural or artificial steroid or thyroid 00:34:58.280 --> 00:35:04.650 align:middle line:84% like which is a steroid like compound, 00:35:04.650 --> 00:35:07.540 align:middle line:84% and it binds to a target nucleic acid. 00:35:07.540 --> 00:35:10.470 align:middle line:84% And both the nucleic acid and the small molecule 00:35:10.470 --> 00:35:12.480 align:middle line:90% have potential for crosstalk. 00:35:12.480 --> 00:35:15.580 align:middle line:84% And here is the nucleic acid part of the story. 00:35:15.580 --> 00:35:17.670 align:middle line:84% And in the next slide will show the small molecule 00:35:17.670 --> 00:35:18.690 align:middle line:90% part of the story. 00:35:18.690 --> 00:35:22.530 align:middle line:84% But the nucleic acid part, you have two protein domains 00:35:22.530 --> 00:35:23.760 align:middle line:90% similar to one another. 00:35:23.760 --> 00:35:25.830 align:middle line:84% This is another example of the symmetry 00:35:25.830 --> 00:35:28.560 align:middle line:84% that we started this talk and ended last talk with. 00:35:28.560 --> 00:35:32.340 align:middle line:84% The symmetry here is, you have these two that 00:35:32.340 --> 00:35:35.790 align:middle line:84% can be direct or inverted repeats, 00:35:35.790 --> 00:35:38.410 align:middle line:84% separated by little spacers here. 00:35:38.410 --> 00:35:40.980 align:middle line:84% So the DNA is in yellow and the little spacers 00:35:40.980 --> 00:35:45.060 align:middle line:90% are in the gray and CPK colors. 00:35:45.060 --> 00:35:47.700 align:middle line:84% And the protein domains are in green and white, 00:35:47.700 --> 00:35:49.620 align:middle line:84% where the green and white are structurally 00:35:49.620 --> 00:35:50.920 align:middle line:90% similar to one another. 00:35:50.920 --> 00:35:54.510 align:middle line:84% It's hard for you appreciate them going around like that. 00:35:54.510 --> 00:36:01.980 align:middle line:84% This is to emphasize that the direct or inverted repeat here. 00:36:01.980 --> 00:36:04.380 align:middle line:90% Now that's the DNA interaction. 00:36:04.380 --> 00:36:07.320 align:middle line:90% And this is the ligand binding. 00:36:07.320 --> 00:36:12.870 align:middle line:84% You can see the estradiol is the small, yellow ligand. 00:36:12.870 --> 00:36:17.500 align:middle line:84% And the tamoxifen, which is the larger ligand. 00:36:17.500 --> 00:36:23.480 align:middle line:84% This is something that's important in treating 00:36:23.480 --> 00:36:28.250 align:middle line:84% breast cancer that might be responsive to estrogen 00:36:28.250 --> 00:36:32.000 align:middle line:90% binding drugs. 00:36:32.000 --> 00:36:34.870 align:middle line:84% So this is the part of the protein that has 00:36:34.870 --> 00:36:37.200 align:middle line:90% two parts, or three parts here. 00:36:37.200 --> 00:36:38.680 align:middle line:90% That's the binding domain. 00:36:38.680 --> 00:36:40.780 align:middle line:84% The little red thing is an activator peptide 00:36:40.780 --> 00:36:44.200 align:middle line:84% and then there's the DNA binding component. 00:36:44.200 --> 00:36:46.690 align:middle line:84% Now what are the crosstalk we have here? 00:36:46.690 --> 00:36:52.810 align:middle line:84% You can see that these wide variety 00:36:52.810 --> 00:36:58.570 align:middle line:84% of different steroid-like protein binding domains, 00:36:58.570 --> 00:37:04.590 align:middle line:84% they bind a vitamin D3, retinoic acids 00:37:04.590 --> 00:37:10.110 align:middle line:84% such as those that occur in developmental processes 00:37:10.110 --> 00:37:11.430 align:middle line:90% and in vision. 00:37:11.430 --> 00:37:13.830 align:middle line:84% Thyroid, which regulates our metabolism. 00:37:13.830 --> 00:37:17.760 align:middle line:84% And estrogen, testosterone, and so forth. 00:37:17.760 --> 00:37:20.730 align:middle line:84% All of these things have fairly similar small molecule 00:37:20.730 --> 00:37:21.720 align:middle line:90% binding sites. 00:37:21.720 --> 00:37:24.840 align:middle line:84% And the DNA sequences they bind are these half sites, which 00:37:24.840 --> 00:37:26.910 align:middle line:84% are very closely conserved in all 00:37:26.910 --> 00:37:28.870 align:middle line:90% the members of this big family. 00:37:28.870 --> 00:37:31.710 align:middle line:84% And one of the main differences is the distance 00:37:31.710 --> 00:37:33.390 align:middle line:90% between these can vary. 00:37:33.390 --> 00:37:35.190 align:middle line:84% And the distance here is indicated 00:37:35.190 --> 00:37:38.850 align:middle line:84% on the far left hand lower left here. dR3, 00:37:38.850 --> 00:37:41.280 align:middle line:84% means direct repeat with three nucleotides 00:37:41.280 --> 00:37:44.400 align:middle line:84% in between those two halves sides. 00:37:44.400 --> 00:37:48.000 align:middle line:84% IR0 means an inverted repeat with 0 nucleotides 00:37:48.000 --> 00:37:49.410 align:middle line:90% between the half sites. 00:37:49.410 --> 00:37:52.560 align:middle line:84% The R15 is direct repeat with 15 nucleotides, and so on. 00:37:52.560 --> 00:37:55.020 align:middle line:84% And you see each family member has 00:37:55.020 --> 00:37:57.252 align:middle line:84% a distinct ligand and nucleic acid. 00:37:57.252 --> 00:37:59.460 align:middle line:84% Although, there's a lot of similarity of the ligands, 00:37:59.460 --> 00:38:01.980 align:middle line:84% a lot of similarity of the nucleic acids. 00:38:01.980 --> 00:38:04.890 align:middle line:84% How do we-- last line of this slide 61-- 00:38:04.890 --> 00:38:08.380 align:middle line:84% target one member of this protein family or other protein 00:38:08.380 --> 00:38:08.880 align:middle line:90% families? 00:38:08.880 --> 00:38:12.150 align:middle line:90% 00:38:12.150 --> 00:38:15.660 align:middle line:84% In some cases, you will have complete artistic control, 00:38:15.660 --> 00:38:19.920 align:middle line:84% not only on the small molecule, but of the protein itself. 00:38:19.920 --> 00:38:23.220 align:middle line:84% If you have a small molecule that looks like ATP, 00:38:23.220 --> 00:38:28.500 align:middle line:84% you can inhibit all sorts of ATP binding proteins. 00:38:28.500 --> 00:38:31.510 align:middle line:84% If you're lucky, you can inhibit a specific class of ATP binding 00:38:31.510 --> 00:38:32.010 align:middle line:90% proteins. 00:38:32.010 --> 00:38:35.520 align:middle line:84% But knocking out a particular member of a class is hard. 00:38:35.520 --> 00:38:38.730 align:middle line:84% And you can see here on the right hand side, these three 00:38:38.730 --> 00:38:39.930 align:middle line:90% chemical structures. 00:38:39.930 --> 00:38:45.960 align:middle line:84% The adenine part are these five and six member fused rings. 00:38:45.960 --> 00:38:49.420 align:middle line:84% And attach to them are the side chains. 00:38:49.420 --> 00:38:51.660 align:middle line:84% The first one, all black structure, 00:38:51.660 --> 00:38:56.010 align:middle line:84% is a known inhibitor of protein kinases in general. 00:38:56.010 --> 00:39:00.030 align:middle line:84% And the red additions are how to make that a little bit bulkier 00:39:00.030 --> 00:39:04.680 align:middle line:84% so that it will no longer bind to protein kinases in general. 00:39:04.680 --> 00:39:07.875 align:middle line:84% Now why would you want to make this inhibitor not bind 00:39:07.875 --> 00:39:08.670 align:middle line:90% to protein kinases? 00:39:08.670 --> 00:39:11.130 align:middle line:84% Well now, if it doesn't bind any protein kinase 00:39:11.130 --> 00:39:13.650 align:middle line:84% very well because it's too bulky, it doesn't fit anymore, 00:39:13.650 --> 00:39:16.770 align:middle line:84% then you can carve an amino acid out 00:39:16.770 --> 00:39:25.380 align:middle line:84% of one of the protein kinases by doing homologous recombination 00:39:25.380 --> 00:39:29.700 align:middle line:84% or transgenic mutating that particular-- the nucleic acid 00:39:29.700 --> 00:39:31.110 align:middle line:90% encoding that gene. 00:39:31.110 --> 00:39:33.780 align:middle line:84% And you will have this ability to manipulate 00:39:33.780 --> 00:39:39.027 align:middle line:84% both the chemical and the protein target in cases where-- 00:39:39.027 --> 00:39:40.860 align:middle line:84% as we'll get to in the last three lectures-- 00:39:40.860 --> 00:39:43.830 align:middle line:84% where we're analyzing systems Biology networks, 00:39:43.830 --> 00:39:47.550 align:middle line:84% you want to be able to target a particular protein at a time 00:39:47.550 --> 00:39:51.930 align:middle line:84% by having a known ligand protein interaction where you minimize 00:39:51.930 --> 00:39:54.990 align:middle line:84% the cross plot by engineering the specific interaction. 00:39:54.990 --> 00:39:58.110 align:middle line:84% You start with a specific interaction for the class 00:39:58.110 --> 00:40:00.570 align:middle line:84% and then you engineer it so it hits one of them. 00:40:00.570 --> 00:40:02.280 align:middle line:84% So that way you can do a time course, 00:40:02.280 --> 00:40:05.100 align:middle line:84% say, just knocking out that particular protein 00:40:05.100 --> 00:40:07.710 align:middle line:90% quickly or letting it come back. 00:40:07.710 --> 00:40:10.972 align:middle line:84% And this shows the results down at the bottom here. 00:40:10.972 --> 00:40:12.930 align:middle line:84% You start out with these two different kinases. 00:40:12.930 --> 00:40:16.620 align:middle line:90% 00:40:16.620 --> 00:40:20.340 align:middle line:90% CDK2, involved in cell cycling. 00:40:20.340 --> 00:40:25.590 align:middle line:90% And chem kinase. 00:40:25.590 --> 00:40:31.320 align:middle line:84% Two, both of these would bind to the original black inhibitor. 00:40:31.320 --> 00:40:35.100 align:middle line:84% And hence, there would be significant crosstalk. 00:40:35.100 --> 00:40:40.530 align:middle line:84% And here, the interfering dosages in micromolars 00:40:40.530 --> 00:40:42.450 align:middle line:84% are shown in the three columns here 00:40:42.450 --> 00:40:43.980 align:middle line:84% for the three different compounds, 00:40:43.980 --> 00:40:45.400 align:middle line:90% run underneath each of them. 00:40:45.400 --> 00:40:53.820 align:middle line:84% And you can see that the lower the number, 00:40:53.820 --> 00:40:54.920 align:middle line:90% the better it binds. 00:40:54.920 --> 00:40:57.080 align:middle line:90% So when you take these-- 00:40:57.080 --> 00:41:02.540 align:middle line:84% if you take these now binding pockets and carve them out, 00:41:02.540 --> 00:41:04.040 align:middle line:84% that's what the little wedge cut out 00:41:04.040 --> 00:41:10.670 align:middle line:84% for the two lowest ones, CDK2, derivative AS1, and chem 00:41:10.670 --> 00:41:14.030 align:middle line:90% kinase 2, AS1 derivative. 00:41:14.030 --> 00:41:16.130 align:middle line:84% Now that you can see, they have a much improved 00:41:16.130 --> 00:41:20.870 align:middle line:84% binding constant to even the bulkiest derivatives. 00:41:20.870 --> 00:41:26.360 align:middle line:84% And this is mediated by a threonine or phenylalanine 00:41:26.360 --> 00:41:28.730 align:middle line:84% at position 38 is changed to a glycine. 00:41:28.730 --> 00:41:30.860 align:middle line:84% Glycine is obviously smaller than a threonine 00:41:30.860 --> 00:41:32.720 align:middle line:84% or phenylalanine at that position 00:41:32.720 --> 00:41:37.910 align:middle line:84% and makes room for the drug just in the same way 00:41:37.910 --> 00:41:40.250 align:middle line:84% that changing the tyrosine to a phenylalanine 00:41:40.250 --> 00:41:42.440 align:middle line:84% made room for the dideoxy terminator 00:41:42.440 --> 00:41:44.790 align:middle line:90% in the earlier example. 00:41:44.790 --> 00:41:49.160 align:middle line:84% So in summary, we have talked about protein three dimensional 00:41:49.160 --> 00:41:52.640 align:middle line:84% structure and how we can program proteins, basically. 00:41:52.640 --> 00:41:56.120 align:middle line:84% How we can use bits of proteins that we may not 00:41:56.120 --> 00:41:57.980 align:middle line:90% be able to predict, a priori-- 00:41:57.980 --> 00:42:00.920 align:middle line:84% from scratch-- how we get from a sequence to a ligand. 00:42:00.920 --> 00:42:03.110 align:middle line:84% But we can take parts that we know 00:42:03.110 --> 00:42:05.240 align:middle line:84% and rearrange them in interesting combinations. 00:42:05.240 --> 00:42:07.910 align:middle line:84% We can build up databases of binding constants 00:42:07.910 --> 00:42:11.570 align:middle line:84% to combinations of combinatorial libraries of nucleic acids, 00:42:11.570 --> 00:42:13.820 align:middle line:90% of peptides, of small molecules. 00:42:13.820 --> 00:42:17.240 align:middle line:84% And we can put these together in novel combinations that 00:42:17.240 --> 00:42:20.450 align:middle line:84% allow us to do network analysis and ask 00:42:20.450 --> 00:42:26.180 align:middle line:90% what protein does what event. 00:42:26.180 --> 00:42:27.770 align:middle line:90% So thank you. 00:42:27.770 --> 00:42:29.650 align:middle line:90% Until next time. 00:42:29.650 --> 00:42:32.000 align:middle line:90%