The path of collapse of the viroid ribozyme or news from fronts using RNAInSpace software was obtained.
A couple of months ago, I talked about approximate results in the task of folding RNA. Let me remind you that you want to roll up the viroid ribozyme NC_003540 of the organism Chrysanthemum chlorotic mottle viroid, whose tertiary structure is unknown.
And so it happened - the ribozyme curled up! ( All existing hydrogen bonds are formed in it )
We look at its final state, and under the cut there is still its folding trajectory, and we also summarize.
')
We look at its folding path
And forget about biology at home, as well as physics, chemistry ...
The approach by which I obtained this folding trajectory will be called the cybernetic-geometric approach . He essentially on the formulation of the problem differs from biological approaches. In this approach, it is considered that the problem of folding proteins and RNA is not a problem of biology, not physics, not chemistry, not any intersection of them (for example, biophysics, biochemistry). This is a purely cybernetic task.
Why? - you ask.
This approach proceeds from the postulate: in order for a protein or RNA to adopt the correct native structure that exists in reality, it is only necessary to take into account the formation of hydrogen bonds, as well as some geometrical positions between the atoms of the nucleotides.
A disclaimer in this way is that this approach does not guarantee an identical match between the model and the real native state. Moreover, in this approach, the presence of conformity is secondary, about the same as the presence of natural intelligence in the field of artificial intelligence.
And the solution methods are cybernetic, respectively, partly from the field of artificial intelligence.
At the moment, the method is based on a general statement within the framework of the problem of game theory, and the application of a number of heuristics. Combining a series of heuristics and finding the exact method is associated with solving the problem I called the “two teacher problem”, outlined in the Summary of the problem “two or more teachers” and the subjective opinion about the AI ​​community and the Concept of structural adaptation and an introduction to the “pure generalization” . From a certain angle, this problem is connected with the "problem of invariant representation in artificial neural networks."
Now, after some experience, it can be stated where and in what my opponents and I were previously wrong and inaccurate.
1. From proteins to RNA is just a popular introduction to biological terminology, there is not a lot of it and this is enough to understand what is being discussed.
2. Mat. criteria are some of the nuances inherent in the task of folding RNA. Required for a deeper understanding and attempts to collapse something itself. But already here a dialogue began, the essence of which just indicates the direction of the approach. The article talks about what characteristics you need to rely on in order to set the suitability function (and what it is that we chewed here ). So biophysicists love to proceed from physical parameters. So this is completely unnecessary - it is enough to describe the geometric conditions inherent in this task. Dealing with physical parameters is very costly - unnecessarily complex calculations, which are still not accurate and can not recreate the real biological environment. It is all the same to calculate the physical conditions in the cell, that if in artificial intelligence we had calculated all the nerve impulses of the brain, we would still get a gross error. Even there, an estimate is given that turning one angle by 0.1 degrees may allow one to turn on the other angle , it is specified 0.2 degrees enough.
3. How to reduce the number of turns the chain? - here are just described / criticized current approaches in solving problems of folding. It is shown that not only do we have errors in the calculation of physical conditions, we are also looking for something that has fallen by the Monte Carlo method, or, in the best case, by the annealing method. It’s like playing chess at random. It also describes a more modern approach from Rhiju Das and David Baker, but alas, we can now say that it is redundant and difficult to calculate.
4. How to evaluate the folding of single-stranded RNA? - here the geometrical characteristics revealed by us are combined to create a fitness function.
5. The limitation of optimizing methods in games with an opponent and without - a clear explanation of why a number of methods that use chance to search for new states are unsuitable for this task.
6. One fundamental problem - here the first touches to the description of the problem of “teaching two teachers” in this task. We also discussed the possibility of using “inverse kinematics” (which is used to calculate joints in robotics), but all this is terribly slow. Accordingly, the developed method can be applied not only to the problem of folding RNA, but also to improve the solution of problems from the field of robotics. By the way, folding is akin to solving the problem of “parking a car” in every single case of contact between two nucleotides.
7. Introduction to folding multi-coiled RNAs - the basics of understanding why the helix is ​​easy to fold, and two difficult. Here are the basics of the correct model.
8. RNAInSpace - software for semi-automatic RNA construction in space is not a good demo.
9. The history of a single reengineering or RNAInSpace v.1.3. Demo - a good demo. Trying the basics! I have almost no feedback :), help ...
10. Collapsing bio-calculations. Again in plain language about the resulting folding model — the folding model described here is false . Well, not so completely - but still. It was just checked, and it turned out that here it is a set of nonsense:
The initial position must be chaotic and elongated. Stacking is important only in certain places where there are no hydrogen bonds. In general, stacking is just additional geometric relationships.
What is the feature?
There are three points why I managed to wind up here, but before that I did not succeed.
1. Increased the speed of folding of individual spirals just up to several hours (versus earlier days). The consequence of a set of reasons: they abandoned the Rhiju Das approach, when the angles are selected from the base, now only a search in the vicinity of ± 5 degrees from the current position. It is important to first go through the search in increments of 10 degrees, and then 0.2 degrees. The previously used step 0.1 degrees is small and superfluous. It is enough to change 6 corners in the main chain and one angle for the side chain, i.e. no change in sugar conformation.
2. Used to change the angles manually. It is important here that it is very difficult to make “subtle” movements manually. to form hydrogen bonds manually is not realistic. But there was one biological moment - this ribozyme is self-cut into two parts. And this is exactly what the trajectory of the folding showed at one time, when folding it is necessary to “cross the two ends”, simulating the cutting. Also manually intervened for another non-biological reason. I do not yet have a program for getting out of forbidden states - when unacceptable covalent bonds are formed (in other words, atoms collide - they fit to unacceptable distances). Therefore, we exit manually. But such manual interventions are relatively few 10 pieces, for 20,000 automatic moves. The truth still needs 100 times to control the direction of the automatic search.
3. In the final stages, it is important that part of the chain should allow the formation of forbidden covalent bonds. This is a modeling problem — the calculated motion trajectory cannot be with the absence of elastic collisions, and it is terribly expensive to count collisions. Therefore, while turning some parts, collisions in others can be ignored.
The main thing is that the described approach (in short, the truth, because I wanted to be quicker to inform, and besides, I would be tired of more detailed descriptions) - it allows you to minimize multiple-stranded RNA. And this may be small - but an achievement. Such automation, which is simple in fact and even if it is not fully automated, no one has yet. The maximum approximation is in the scientific-game FoldIt , but there everything is much more difficult for being able to collapse.
Further, I think to test for tRNA - there are already three helices in a fairly complex combination.
PS I apologize for some confusion ...
And so it happened - the ribozyme curled up! ( All existing hydrogen bonds are formed in it )
We look at its final state, and under the cut there is still its folding trajectory, and we also summarize.
')
We look at its folding path
Disclaimer
And forget about biology at home, as well as physics, chemistry ...
The approach by which I obtained this folding trajectory will be called the cybernetic-geometric approach . He essentially on the formulation of the problem differs from biological approaches. In this approach, it is considered that the problem of folding proteins and RNA is not a problem of biology, not physics, not chemistry, not any intersection of them (for example, biophysics, biochemistry). This is a purely cybernetic task.
Why? - you ask.
This approach proceeds from the postulate: in order for a protein or RNA to adopt the correct native structure that exists in reality, it is only necessary to take into account the formation of hydrogen bonds, as well as some geometrical positions between the atoms of the nucleotides.
A disclaimer in this way is that this approach does not guarantee an identical match between the model and the real native state. Moreover, in this approach, the presence of conformity is secondary, about the same as the presence of natural intelligence in the field of artificial intelligence.
And the solution methods are cybernetic, respectively, partly from the field of artificial intelligence.
At the moment, the method is based on a general statement within the framework of the problem of game theory, and the application of a number of heuristics. Combining a series of heuristics and finding the exact method is associated with solving the problem I called the “two teacher problem”, outlined in the Summary of the problem “two or more teachers” and the subjective opinion about the AI ​​community and the Concept of structural adaptation and an introduction to the “pure generalization” . From a certain angle, this problem is connected with the "problem of invariant representation in artificial neural networks."
Run through the previous articles
Now, after some experience, it can be stated where and in what my opponents and I were previously wrong and inaccurate.
1. From proteins to RNA is just a popular introduction to biological terminology, there is not a lot of it and this is enough to understand what is being discussed.
2. Mat. criteria are some of the nuances inherent in the task of folding RNA. Required for a deeper understanding and attempts to collapse something itself. But already here a dialogue began, the essence of which just indicates the direction of the approach. The article talks about what characteristics you need to rely on in order to set the suitability function (and what it is that we chewed here ). So biophysicists love to proceed from physical parameters. So this is completely unnecessary - it is enough to describe the geometric conditions inherent in this task. Dealing with physical parameters is very costly - unnecessarily complex calculations, which are still not accurate and can not recreate the real biological environment. It is all the same to calculate the physical conditions in the cell, that if in artificial intelligence we had calculated all the nerve impulses of the brain, we would still get a gross error. Even there, an estimate is given that turning one angle by 0.1 degrees may allow one to turn on the other angle , it is specified 0.2 degrees enough.
3. How to reduce the number of turns the chain? - here are just described / criticized current approaches in solving problems of folding. It is shown that not only do we have errors in the calculation of physical conditions, we are also looking for something that has fallen by the Monte Carlo method, or, in the best case, by the annealing method. It’s like playing chess at random. It also describes a more modern approach from Rhiju Das and David Baker, but alas, we can now say that it is redundant and difficult to calculate.
4. How to evaluate the folding of single-stranded RNA? - here the geometrical characteristics revealed by us are combined to create a fitness function.
5. The limitation of optimizing methods in games with an opponent and without - a clear explanation of why a number of methods that use chance to search for new states are unsuitable for this task.
6. One fundamental problem - here the first touches to the description of the problem of “teaching two teachers” in this task. We also discussed the possibility of using “inverse kinematics” (which is used to calculate joints in robotics), but all this is terribly slow. Accordingly, the developed method can be applied not only to the problem of folding RNA, but also to improve the solution of problems from the field of robotics. By the way, folding is akin to solving the problem of “parking a car” in every single case of contact between two nucleotides.
7. Introduction to folding multi-coiled RNAs - the basics of understanding why the helix is ​​easy to fold, and two difficult. Here are the basics of the correct model.
8. RNAInSpace - software for semi-automatic RNA construction in space is not a good demo.
9. The history of a single reengineering or RNAInSpace v.1.3. Demo - a good demo. Trying the basics! I have almost no feedback :), help ...
10. Collapsing bio-calculations. Again in plain language about the resulting folding model — the folding model described here is false . Well, not so completely - but still. It was just checked, and it turned out that here it is a set of nonsense:
But there are also so-called interaction stacking ... So if RNA appears gradually, then before the first hydrogen bond can form, at least twenty nucleotides appear. ... And before, I thought that they were just stretched into a chain ... Stacking turns out to be important initial positioning
The initial position must be chaotic and elongated. Stacking is important only in certain places where there are no hydrogen bonds. In general, stacking is just additional geometric relationships.
What is the feature?
There are three points why I managed to wind up here, but before that I did not succeed.
1. Increased the speed of folding of individual spirals just up to several hours (versus earlier days). The consequence of a set of reasons: they abandoned the Rhiju Das approach, when the angles are selected from the base, now only a search in the vicinity of ± 5 degrees from the current position. It is important to first go through the search in increments of 10 degrees, and then 0.2 degrees. The previously used step 0.1 degrees is small and superfluous. It is enough to change 6 corners in the main chain and one angle for the side chain, i.e. no change in sugar conformation.
2. Used to change the angles manually. It is important here that it is very difficult to make “subtle” movements manually. to form hydrogen bonds manually is not realistic. But there was one biological moment - this ribozyme is self-cut into two parts. And this is exactly what the trajectory of the folding showed at one time, when folding it is necessary to “cross the two ends”, simulating the cutting. Also manually intervened for another non-biological reason. I do not yet have a program for getting out of forbidden states - when unacceptable covalent bonds are formed (in other words, atoms collide - they fit to unacceptable distances). Therefore, we exit manually. But such manual interventions are relatively few 10 pieces, for 20,000 automatic moves. The truth still needs 100 times to control the direction of the automatic search.
3. In the final stages, it is important that part of the chain should allow the formation of forbidden covalent bonds. This is a modeling problem — the calculated motion trajectory cannot be with the absence of elastic collisions, and it is terribly expensive to count collisions. Therefore, while turning some parts, collisions in others can be ignored.
What's next?
The main thing is that the described approach (in short, the truth, because I wanted to be quicker to inform, and besides, I would be tired of more detailed descriptions) - it allows you to minimize multiple-stranded RNA. And this may be small - but an achievement. Such automation, which is simple in fact and even if it is not fully automated, no one has yet. The maximum approximation is in the scientific-game FoldIt , but there everything is much more difficult for being able to collapse.
Further, I think to test for tRNA - there are already three helices in a fairly complex combination.
PS I apologize for some confusion ...
Source: https://habr.com/ru/post/149449/
All Articles