Video 2-4: Gene detection

0:03 In the next section, we'll look at how you could detect in this very long remember three billion bases for a human genome. And there's there's species with much larger genomes. We're no exception. Uh we're neither very short nor very long in terms of the the genomic sequence. But three billion is quite a lot. How do you find out um where there are genes or uh how uh uh what what they what they do? Uh and so I want to zoom in a little bit on on this specific task. Suppose for some reason you've already managed to get the full genome like in our case we have this since 20 years. It was co-developed. But suppose you just have the sequence of the whole genome and you don't know what to do with it. How could you uh try to identify these uh protein coding regions? That's by and large what a gene is for us. There's a a region where this genetic code gives you a a protein.

1:04 And there's a few things as as almost always in uh in uh bioinformatics. There's two options which uh uh one is you just take this data as I as I mentioned and you try to do something that's intrinsic just to that piece of data that piece of information. I'll discuss the options in a second. The other option is you somehow involve other knowledge about the biology and collect additional information maybe additional uh uh experiments in the lab and so on. Almost all real results in research are a combination of these. Uh but it sometimes helps to keep keep things apart. We'll focus on this app initio uh gene detection where we're just taking the the letters and try to figure out figure out what we can say from that. And the first first one that we'll focus on is this uh start and stop code on um meaning in the code. Maybe it's uh worth going briefly back. So you notice um there's one start code on which by and large uh you need this in the DNA sequence to start the translation of a gene or this translation of a protein.

2:20 And then there's one of these terminating the translation a sequence in the in the DNA that begins with one of uh with with a single start codon and ends with one of these and in particular doesn't have any other occurrences of the stop codeons. That's called an open reading frame. And you you should think of this like a this is kind of opening the book and this is closing it shut. So uh trying to read a gene immediately stops when one of these uh occurs. So you can't keep going. Uh there's another thing that usually to start the translation. Remember this fat uh brown blob that uh um was kicking off the translation in the video. the uh RNA polymerase this has to somehow attach to the DNA and usually there's some if you think about the DNA strand and say here's the start code on and then you keep reading and there's an end stop codon somewhere somewhere here so usually there's a little upstream from the start code on some specific sequence that uh allows this um the RNA polymerase to attach and stick there. So you can also try to find these. Unfortunately, these are not always the same. So we'll we'll come back to these um in the latter part of of the unit. Um they're the tricky bit is that they may be a different sequence in different uh species.

4:05 There's a third one that's a little more subtle is one of the things that um didn't make it into the module in the end. Uh but there's um if you look at it statistically there's a subtle difference in how often the different letters occur in a protein coding sequence and outside of these. It's usually too subtle to see with a naked eye but you can train a specific stochastic or a statistical model to notice the difference. It's not an exact uh thing. So it's only um a probability scheme, but you can use these that's essentially uh hidden markoff models that you can use for that if you've maybe heard the term, who knows? Um they can essentially give you a probability whether you're in the gene mode at this point or in the non-gene mode. And they can transition with some probabilities, but it's a very fuzzy business. They'll only give you a hint. it's more likely to be here or more likely to be here.

5:06 Okay. Uh for today we'll focus really on this first part and you could ask yourself is there any hope with just these uh start and stop codons? Uh is there any hope to detect genes in any somewhat reliable way? And what helps us here is that uh evolution is a certain directed force. Uh over long stretches of time, there's always going to be random mutations in the DNA. Uh just because these these bases, I mean, we think of them as four distinct letters, but they're actually consisting of of smaller molecules or smaller atoms. And you can convert from one to the other if you change some of the atoms, which sometimes happens by random chance. And also in the copying process, you can you can sometimes have mistakes even though they're not so likely. So there's there's going to be these random changes in letters. And now two things can happen. Either this uh this essentially destroys the functional part of whatever it was uh modifying. Maybe some gene now encodes for a different protein and this one doesn't function anymore.

6:18 If that's bad enough, if that's fatal for the organism, this mutation will just die out, right? This this is no longer observable because all the creatures that had that mutation are no longer with us. On the other hand, there might be um mutations that don't change things uh and then these might still be visible and there might be parts in the DNA where uh at least the exact function isn't clear. But for many parts, we know there's heavy differences between different species and even individuals of the same species and nothing really happens.

6:57 So, uh it's it's somewhat reasonable to assume that there's parts in the genome where mutations are important to avoid and then these will be highly conserved. Whereas for the other parts, we don't have a reason to assume that. So probably they look some somewhat random because uh the mutations are dictated by random events and if nothing in the evolution stops them from being selected against then they will stay random. So we'll always compare what we see in some specific genome to a null model which is that stretch of DNA is just generated randomly. And if we observe something where the probability to see it in the null model is is very very low, we can take that as a signal that we found something of importance of biological significance one way or another.