Video 8-1: Noncoding RNA

0:03 In this next chapter, we'll look at RNA secondary structures and how to predict the structure if you're only given the sequence. And uh I realized that we haven't talked much about RNA. So we'll uh talk a bit more about the biology quickly and the the usual way with some um some videos and pictures.

0:23 um and then I'll introduce formally what the problem of this prediction um structure prediction problem is and what we can do there. First of all, um we've already talked about RNA in one context at least, namely in the central uh dogma of molecular molecular bi biology, but there was only used as kind of a boring part in the machinery to produce proteins. And here in this in this unit, RNA is kind of a center stage as a first class citizen part of the function in in cells. What is RNA? It stands for ribboucleic acid. It's similar to DNA. Um it's formed of this chain of sugars and the type of sugar is slightly different in the two. um which means they look almost the same often, but the RNA backbone is slightly different. It's a bit more sticky because there's one more of these um hydroxal groups that can form bonds. Uh whatever the chemistry in detail um does for for a different lecture.

1:38 Um for us what's important is that as for DNA there's four different bases that usually occur and that you can use uh to chain together in a molecule. Unlike DNA it's usually just a single strand. So DNA always the two reverse complimentary copies that then twist into the double helix. For RNA usually just a single strand that still does twist. We'll come back to that. uh we've seen the messenger RNA as the the key role in uh protein synthesis where again you may remember this picture is a bit small here uh in the nucleus of of ukareotic cells you have DNA there is a copy created from the DNA and the copy consists of RNA that's the messenger RNA which then travels out to the ribosomes where it's translated to protein but we've talked about that in detail scale. Um, fascinating in its own right. However, this is not all that RNA does. Um, there's many other roles that RNA serves. And there's even I think it's not more than a hypothesis really. Um, but the hypothesis of the the RNA world that potentially RNA was there first and initials life forms when life first evolved consisted just of RNA and kind of the more stable DNA came after I mean invented is the wrong term. I'm not sure it it m randomly formed a bit later. RNA can otherwise uh serve most of the roles. It can preserve information like uh genes. Um it can s can have an active function in biosynthesis and it can perform most functions that we now attribute to proteins.

3:36 So in a way it makes sense that if you had just a single building block it might have been RNA. Okay. Um the video I want to show you is from some uh teacher training program. So it's targeted at slightly different group of people. Uh but I think they have the nicest visuals to quickly grasp what this is about. Uh which is 3D printed structures. Um let's see.

4:12 >> All right. So I want you to feel really comfortable about complimentarity uh because that is going to come up over and over again in this in this module. Now in addition to DNA base pairing um with itself and RNA base pairing with DNA in order to form um an RNA transcript RNA can base pair with itself and DNA a DNA strand can base pair with itself. It's possible if there are stretches of complimentarity in one strand of RNA or DNA for that strand to flip over onto itself and um and bind to itself. Sometimes they call that a hairpin if it just takes a it flops over and binds to itself. That doesn't that isn't always the case. So all RNA is not destined to be protein to be translated into protein. And you probably are already aware of that. You just don't maybe think of it that way. But there's a lot of RNA that is called non-oding RNA, which means it doesn't code for proteins. It is coded to be um a m molecular machine all by itself.

5:38 It's it's going to have secondary and tertiary structure and act like a molecular machine like proteins do. Um, and you know of some of these already. Okay. So, if we look at this, we talk about transfer RNA. Well, transfer RNA is non-coding RNA. It's a molecular machine all by itself. we talk about the ribosome, ribosomal RNA and that's what we have right here. Okay, it is um it is a molecular machine made up of primarily RNA. So how does that work? So we know that proteins based on their sequence fold into a very specific shape that is important for them to carry out the jobs that they have or the job that they have. Um and RNA is is similar. It folds into these very complex shapes uh based on a couple of different factors and one of them is complimentarity that we were talking about before. So base pairing within itself and if you look at a tRNA you can see that it is just one strand of RNA that circles through and then base pairs with itself. And then what it does is it turns into a very um twisted complex 3D shape because of a stickiness of the backbone that we're not going to go into. But the fact that it's ribboucleic acid and not deoxy ribboucleic um nucleic acid makes it have a sticky backbone that will allow it to um contort and stay into these shapes.

7:41 So on this graphic you can see that um that RNA as we know is single stranded and if it is destined to be a transfer RNA um and not a pro translated into a protein um it then can base pair in this kind of clover leaf pattern they call it um because it's got these regions of complimentarity that we were talking about before. And then because it's got that sticky backbone, it then twists into this three-dimensional shape that is very specific for its sequence. So this structure is very specific and very required so that it fits into the ribosome just right so that it can add that amino acid to the end of the growing peptide strand. Right? So that is one uh example of a non-coding RNA.

8:44 Okay. So in addition to transfer RNA, you are also aware of [clears throat] of something called a ribosome and that's what this this purple uh structure is here. But this is what but this >> Okay, I think that's that's enough um for now for getting uh the gist of of many things. I mean, uh, I find the the play area behind her really cool where you can see how you do copies of DNA and so on. Feels like a fun thing to have. U, but, um, apart from that, there was a lot of, uh, a lot of detail already in there that we'll come back to a bit. Uh, I feel seeing it in in real 3D is kind of more helpful, even if it's just a video, than than the animations. I tried for myself. um there's certain things that in the abstract model I'll show you seem plausible but then they don't really work in the three-dimensional reality.

9:48 Okay. So apart from apart from the functions that were mentioned like the transfer RNA that translates from genetic code to proteins and the ribosomes that puzzle proteins together. Uh many other functions are known. Um and one one key property why why bi biologists care about the function apart from predicting sorry care about the structure apart from predicting what the function in the cells is is that um for RNA often the structure is preserved among different species but maybe the sequence isn't. So if you were to look for uh um parts in the the genome that correspond to the same function, they might as kind of the letters in the gen the genome look very different because exactly what uh bases you have is irrelevant if they don't change the shape. So you can make substitutions as long as they're paired in a way that they don't change the structure. Um they don't really affect the function. But if you want to detect similarly working uh molecules in in the genome, you have to is instead of just the sequence, look at the structure. Otherwise, it doesn't make sense.

11:02 Uh that's that's the other reason why people care about predicting the the structure a lot. Okay. Um how do people even know how to make these 3D models? How do we know that they are somewhat realistic? Uh there are lab techniques that you can use to get the 3D structure. I think I mentioned this very briefly when we were at the very beginning of the module talked about protein um sequencing. Remember the the turnpike problem and these things. Uh in the end there I said the much cheaper way is to throw the molecule in a into a mass spectrometer, break it into uh random fragments and then throw it into a mass spectrometer. Uh if you want the actual 3D structure measured, that doesn't work. Um but you can clean up uh the protein or the RNA so that you have it in very pure form. Then it can crystallize and these crystals you can throw in an X-ray and observe the pattern. Um is a very indirect metric uh is a bit like computer tomography. So very um very a lot of uh well indirect reasoning about the structure involved. um needless to say is also expensive and very laborious. So you can do this for some important molecules but people are very uh interested in doing cheaper alternatives.

12:30 Um um unlike for proteins, the alternatives from the machine learning side of things don't seem to work for RNA so well yet. And the main reason is lack of data. Uh we'll we'll come to that a bit uh later. Again, I want to show one one quick animation that shows another problem that after that video kind of as a um disclaimer will ignore. Um and that's the fact that there's not just a single structure usually, but rather um our name. All right. So that we had That's the one. So they uh they tried to measure in real time how these molecules behave and look like. uh and it doesn't immediately find the right structure but rather this this yellow part forms step by step in that uh initially it's uh it's in some random shape and then step by step it forms into the final structure the final shape that it then takes [snorts] uh but you see it still wobbles around and that seems to be just play it again uh that seems to be also something that happens remember these are all essentially in water in your cells um and it's reacting with other things. It might depend on uh how much uh salt and how much uh acid there is in the environment in exactly what shape it forms. So some molecules are also known to flip between different confirmations and that also has function. We'll mostly ignore this and just focus on predicting the single most stable form. Um, in particular for comparing say between different species which RNAs have the same function. That's that's good enough.

14:47 What we'll also look at here because there's not enough data to do much better are denovo approaches which means we're just given the sequence and out of nothing we try to predict what the structure could be. [snorts] So most most of the models will have to somehow just incorporate the chemical knowledge of what confirmations are stable and not use much else.