Assistente AI
Trascrizione
00:00:14Non so. Notes on the schedule of
00:00:19the course so there is class today tomorrow their one be class.
00:00:26Newsweek in with the conference conference
00:00:31and be the last two classes on the third and the heart of June.
00:00:39With tomorrow.
00:00:41Finish the program on the third hour
00:00:45a view of the program so you you will be of the life
00:00:52of questions that so you all the topics and right to highlight what
00:00:59are the important things that it explain during
00:01:02the course and on the first to be a seminar from a league.
00:01:08Alexander Monson, who is an expert on structural sample and
00:01:13so what is on the field.
00:01:21To modo all also introduce how
00:01:26to di final project so today have to finish couple of
00:01:31things before talking about that and
00:01:37that's it for what regards di the special am so today.
00:01:45Al finish di and simulation simulation exploration
00:01:52and cover up of software,
00:01:54which is call for.
00:01:58Interaction generation that is that is
00:02:03important for the final project to model a focus on protein
00:02:09embedding and how they can be used
00:02:12in matrix to solve couple of problems that are
00:02:16very rilevante today so
00:02:20first time to models also the deadline for the second mid-term.
00:02:25So I've seen that many of view
00:02:29that so is the question about that problems.
00:02:37Was super easy am other questions.
00:02:45Body final project exam its it is clear that
00:02:51final exam is about doing project Explain
00:02:56the project to me plus couple of
00:02:58questions about the course program so
00:03:01please be aware that
00:03:05we ask something about the course is not just a project.
00:03:09Okay, am sono questions not.
00:03:18Modified likely di te slides for the molecular dynamics.
00:03:24Just the ones that is not present
00:03:27last just minor things am and I opened.
00:03:33The slides for what regards things Where
00:03:38are you can download am a end of
00:03:44the course al make public or the recordings so if you
00:03:49have thoughts about something you can go back and
00:03:52see di the Recording Lectures.
00:04:00So we were looking
00:04:04dynamics lost stime were able to
00:04:08cover society exploration of the force fields so
00:04:11we can model forces between buttons or we can simulate or
00:04:19proxy and simulate forces and so the concept of energy functions.
00:04:27So where that are functions that describe what is the energy
00:04:32will see that energy and forces are in the connected concepts,
00:04:38of course, and so energy can be calculator
00:04:42from the initial description of the state of the system to
00:04:48wait and functions with something similar ui a so he
00:04:55statistical potential so in case
00:04:58of the energy of a confirmation was.
00:05:09So in that case for instance
00:05:14the PDF potential it is the energy or probability of
00:05:20having like information protein just looking at
00:05:25the distance between specific type of specific type of
00:05:30its number having idrofobico acids distance indicator of having
00:05:40a meno assets in the core so they are close index that
00:05:44they are place in the structure as well
00:05:49having That are about this is
00:05:54that the solution is racing and sorpreso.
00:06:03In this case we are modeling.
00:06:10This time of the total energy of the system,
00:06:15This given by the energy bond structure and
00:06:19the energy of the bond structure
00:06:21of the bond structure is actually di
00:06:24bond structure so model how
00:06:30what is the and you have to outcomes that are connected chemicals
00:06:35when you baristi di optimal equilibrium is
00:06:41e non bond elements or energy is given by the static component and.
00:06:49component.
00:06:59So why this
00:07:04is so model the end of
00:07:08the connected options simulate an armonico simulator or a spring
00:07:14or Slater so you see how we model system with
00:07:19this formula so what equilibrium is
00:07:23this distance at the optimal distance between
00:07:26two nuclei you have the minimum energy and if you deviate from
00:07:31this optimum your energy actions even if even
00:07:38when you get long distance and also when
00:07:41you get short this long distance of course you are.
00:07:46At the DLT distance which is call
00:07:51to action distance where the two options will break apart
00:07:56and energy that there is a specific break di di di bond.
00:08:09Simulation with actually go that further so
00:08:15we don't actually that this is
00:08:17that we are not interested in breaking the structures.
00:08:20We are just interested in capture how cost it is
00:08:25the nations from the optimal optimal so we model is this way.
00:08:33And similar for angels
00:08:35for and so if we want to the end of time. We have to.
00:08:42Take it to account old business between.
00:08:46Connected physical for the interaction
00:08:52actually we have to to evaluate
00:08:54all possible combination and system.
00:08:58In principle any person.
00:09:03Make interaction so we have to
00:09:06evaluate what is the energy contribution.
00:09:11Within a certain so what is the formula in this case you see
00:09:19that after the certain tank energy competition
00:09:24is so that if you are.
00:09:27Distant half diagnosi visiva,
00:09:31but the interesting think about this formula
00:09:35is di the fact that you have the distance
00:09:41to times to numbers di equation and in one case you see the air
00:09:51is the actual business between to nuclear asilo
00:09:56is a fix parameter and you see that the pending weather.
00:10:01This is your own lower then
00:10:05are you have a number that is the one or
00:10:10number one and we are
00:10:15different the pending on the am on the situation.
00:10:20So what is that you have one to one and you have
00:10:27the minimum one is that you have a number that is lower year.
00:10:35In parentesi number one and so.
00:10:41One of the two experience dominance on where you
00:10:46are in the problem of this time is
00:10:52that is for date so you have to calcolate energy for
00:10:58peace and one of the time consuming steps of the times.
00:11:08The other energetic is to this is the formula again. This is.
00:11:19Possible person in your system and you see
00:11:25what is the relationship between the distance and
00:11:29the energy you see here di case to atom with targets and di
00:11:39diagnosi Always negative and of course it is lower when
00:11:46the two charges are more than if you have the same charges,
00:11:52of course the is the opposite so you have the same plot that
00:11:56is over di x axis so if you are the charges that get closer,
00:12:04the energy is that they actually don't like to stay stay close to
00:12:10to each other so the problem of this is that the distance.
00:12:17Which is longer compete to the world is so what is the point.
00:12:26The energy is so simply apply evaluate
00:12:31only within the distance is this distance actually for.
00:12:41Ho letto Static turn the distance is much
00:12:46larger and so you have an energy contribution.
00:12:51Even if two charges are far parte in
00:12:55the system so this is actually di to the bottom.
00:13:06Is it clear other questions actually if
00:13:11you know something about this are
00:13:15actually am expansion or form that is what you expect.
00:13:25Physics applied other context so it's perfect. IT works.
00:13:32Okay, so we we didn't see yet what we can Cliccando qui.
00:13:38This force fields that in second
00:13:42just a few more details There is another problem where simulate
00:13:48system the computer which is di
00:13:53the boundaries boundaries conditions and
00:13:58see that for simulator integrated di the system over time.
00:14:05You could have actually your modules the coordinate
00:14:08the shift or move along
00:14:12your computational space actually what you
00:14:17do is you define founding box so
00:14:22box where your protein is inside and you also
00:14:28defined some cases espliciti modules so your box you will feel.
00:14:36number of the end,
00:14:39of course this is a cost because what we are
00:14:42actually interacting with your protein and you
00:14:45have to calcolate all di un piano di
00:14:50intervento Typekit interaction interaction in water all your
00:14:54own your prophet so the problem of
00:15:01system like this is that you can exit in the box.
00:15:08So you can either define am away where
00:15:12you am you need what is the direction of
00:15:17your molecular so is like it is on on the edge of
00:15:21the box or you have something like it
00:15:25system where when the year your protein exit
00:15:28one it enter the edge on the on the other side of
00:15:32the box Or you can be something where you with
00:15:36your system multiple times and the edges for example.
00:15:41If you have multiple proteins multiple modules
00:15:44they can interact with d with the box that
00:15:48is in the next box di a di the boundary of your simulation.
00:15:58Ok, so the interesting part so we have the fine force fields,
00:16:05so we have the energy given by the state in our case.
00:16:11The state is the distance and of course the parameters that are
00:16:16the associated values di
00:16:19equilibrium distance and so on what we have the energy.
00:16:25We can also Calcolate the force
00:16:28so what is called
00:16:32potential energy and this is actually the formula so before.
00:16:37Sapevi di Oscillator If you want to the force,
00:16:42we can simply a colpi di
00:16:46derivati over di di di this because this one dimensional.
00:16:52This is the formula che so the force is actually indicate.
00:17:01Al far we are from the equilibrium
00:17:06what is the strength that we want to express?
00:17:11Di again di riequilibrio od minimum the minimum
00:17:16and so is it clear this is.
00:17:21One of the main thing so there is
00:17:26the connection between energy and forces and of
00:17:31course problem in our case is that we have
00:17:34to deal with the three dimensional space so we can
00:17:37not simply apply it is on one direction,
00:17:45one dimension, but we have to get it to
00:17:48the three dimensions and when we can di.
00:17:56One man show. The corrispondente.
00:18:00Mathematical tools is gradient
00:18:06is this symbol year of the gradient of.
00:18:14Energy in this is can express the force direction where you
00:18:24should follow to the energy and also it
00:18:29device direction because we are component or one.
00:18:36Direction is always same with component.
00:18:40We could gradient that is different for the three,
00:18:44he is it actually the last
00:18:48where stb should go imaging something like this simple.
00:18:54What the state the position that indicate where you
00:18:59can be energy in the gradient.
00:19:04We will actually see where this is what like to go to
00:19:09many miles so the gradient
00:19:13actually has what is the direction and di.
00:19:19And magnitudo of the force or
00:19:23the displacement would like to to follow to the minimum Marras.
00:19:29Things about the important.
00:19:34Year In realtà the flip data.
00:19:41That is the same thing a force and.
00:19:47So how we can minimis energy.
00:19:51There is nothing different to what you see
00:19:55in your mathematical courses.
00:19:59So what when the first derivative
00:20:04zero identifier station points when the second derivative.
00:20:09If you evaluate second of course you can also am
00:20:14evaluate whether we are on local minimum local maximum or year on.
00:20:22So am again.
00:20:25We have been from one dimension to the three dimension and again be
00:20:35the same thing for
00:20:37the gradient di only difference is the second life
00:20:42is not in my back to have to what is call di.
00:20:51Matrix and local minimum local minimum
00:20:58is when the use of this matrix are.
00:21:05Not about about this matrix completed aim.
00:21:14Of course you di
00:21:17del West of the local minimum is the global minimum and
00:21:24optimization here is not the place to
00:21:29find di set of coordinate that the end,
00:21:34of course question is how you
00:21:39minimax that energy and
00:21:42that's the objective of the molecular mechanics.
00:21:46So mind that there is difference between
00:21:50molecular mechanics and
00:21:51molecular dynamics molecular mechanics about
00:21:55this following the gradient as so following status confirmation.
00:22:12This is why?
00:22:17So start from your information you con più
00:22:23di di di energy and so energy we know.
00:22:30What we should move Atlas?
00:22:36Energy Force field so the force field so
00:22:40before is used to this potential energy.
00:22:46Using a mathematics and then you can can di
00:22:53updated positions by simply playing di Ehm.
00:23:02Ehm at the gradient displacement multipli
00:23:07actually The Guardian di initial position at
00:23:12the position and you see we use usually parameter
00:23:17to a tutti Un what is want to.
00:23:23Check for convergence so you are happy with
00:23:28your new Orleans and if not you you need
00:23:34is to this is that we can you
00:23:39can do and you update the position by integrated.
00:23:47So. Di question is now you need for convergence and ideas.
00:23:58When you will be Satispay.
00:24:01Da options.
00:24:07Non quotate slides Start.
00:24:13Up difference. Ciao a.
00:24:18So if you apply
00:24:23the gradient di exact is identica to the previous the coordinate.
00:24:30Actually if the structure doesn't move.
00:24:34You know how to do is not force moving out of more than one.
00:24:45And the rest connected that's the same teso a.
00:24:54Said you can also be energy so the energy is the same.
00:25:01This is the final and the previous one of the same.
00:25:06And. The position confirmation is the same.
00:25:15So The problem of this approach.
00:25:22Is that it is on the initial geometry
00:25:27and I think about that is follows the past.
00:25:33This is actually following the the natural path and only capture
00:25:42a local local minimum actually able to Alessio and or you
00:25:49employment something smart to sample different initial position in
00:25:54order to increase probability of finding di a global.
00:26:00Anyway, this is way to what you think is di diagnosi,
00:26:09of course the energy is is less because we are moving
00:26:13towards less energetic Start but that
00:26:19is a way to to to remember
00:26:21that usually when you do this it means you
00:26:25del Testaccio so if you think about
00:26:28that we are energy is something were happens
00:26:32an positions that is not optimal the sort
00:26:37of your energy they are aim.
00:26:43In atto dal.
00:26:46Dispositivo di there are conflicts that could be classes and so on.
00:26:53When you apply this is your relax your finding information that.
00:27:00Is actually the the things that you see when you simulate
00:27:07a PDP structure even sperimenta solution when Meglio.
00:27:15Force field to optimus this you will see that the protein
00:27:20a little bit infilate su
00:27:23the positions buttons becomes the structure because.
00:27:29This is because this experimental conditions when you from
00:27:34the crystal buttons particular when the air force
00:27:39the form something without solve and the protein is to
00:27:43pay to this possible amount space
00:27:48when you force field which instead.
00:27:53Fisiologica or solve conditions the structure immediately
00:27:59becomes a little bit more longer on the move bit.
00:28:06Okay, so simulacra names is
00:28:11different thing Dynamics uses the principle of molecular mechanics,
00:28:18but for molecular mechanics.
00:28:21There is something we get static snapshot of
00:28:25information solutions driving position
00:28:30of the minimum energy molecular dynamics.
00:28:34What we apply is classical mechanics,
00:28:39so force mass acceleration
00:28:45and we actually am integrate over time.
00:28:49So di the year is forward
00:28:56and that we can write di force
00:28:59mass by acceleration and acceleration,
00:29:04second time of displacement and you can express.
00:29:10IT is And you can also.
00:29:17Write on this way so we have.
00:29:22The time of the mask by the lost city and must by
00:29:28velocity is the momentum momentum is the must multiplayer velocity.
00:29:36I think about this must be
00:29:38velocity is actually that something that incorporated.
00:29:43Again the magnitudo and speed this is to think about this is
00:29:51the speed anatomy moving towards giving
00:29:55directions to the difference with molecular dynamic simulation.
00:30:00Molecular Mechanics is that you are not
00:30:03testing how molecular is going to the minimum.
00:30:09Ma di testing How molecular eyes over time so your not interested
00:30:17in looking how it's minimal where you are interesting moves?
00:30:29So connection between this formula and force so
00:30:34before where is defined as the negative.
00:30:40First time out of the time.
00:30:46And so the derivatives of momentum over time is the negative.
00:30:56Derivative of the potential over the place
00:31:01so keeping this in mind we can the voice a new.
00:31:07Age. Lo. So we
00:31:14shall conditions initial definition again
00:31:20position and momentum so
00:31:22the difference here is that we don't calcolate.
00:31:26We have momentum so we have to the fine also what is
00:31:30the last of us so initial velocity.
00:31:33We have to provide what is parameters the beginning,
00:31:40we calcolate potential energy and gradient.
00:31:45We update the gradient the position
00:31:49instead of aim playing the gradient
00:31:55we updated using the momentum so information
00:32:00about velocity is to be the final mission.
00:32:07qui. Update the momentum so we exploit the connection between
00:32:13the momentum and the gradient and so we can
00:32:16updated momentum thanks to the gradient.
00:32:20We have calcolate before so what you are.
00:32:27Testing where you are actually when we
00:32:32go apply that momentum and immediately after you
00:32:38have the momentum has changed in
00:32:41comparison with that displacement that we have just so
00:32:46this is actually capture am what was
00:32:52the effect of that velocity or
00:32:54that displacement and so it updates not only the position,
00:32:58but it updates also the velocity buttons if your moving in
00:33:04the Conversion velocity is
00:33:08better to get back and move di iPhone di diapositive back.
00:33:16Again until you is what you want to you decide that you di
00:33:27the final part of course
00:33:30di includes the displacement and increase in the momentum,
00:33:38of course momentum, but also the friends on
00:33:43the time step with the final
00:33:47because this point you see with multiply values by.
00:33:53Amount of time so the time step you see is
00:33:59the critical factor here because this Small time frame
00:34:04the times that the amount
00:34:08of looks we have to keep it the are the times
00:34:14that the larger probability of
00:34:18the others and finding a displacement that
00:34:22is actually a creating conflicts or is not the year.
00:34:31The point, of course initial condition.
00:34:38Definition of the times.
00:34:44So years of the difference between
00:34:49the two molecular dynamics versus minimis.
00:34:54So. And just to think about
00:35:02the time so there is a time step usually very short, but also am.
00:35:13AM di tipo if you want see anything you see the moves a little
00:35:23bit what you want to capture for
00:35:27interaction a collision or or like that it is to long.
00:35:34There is Bischoff am
00:35:37using so you to find something that is reasonable
00:35:44another problem that is how long I
00:35:47should simulation and of course again
00:35:50is about the amount of
00:35:53the sources that you have the end, of course.
00:35:57What is the process that you want to to capture simulation long
00:36:03even if you have a short protein small protein
00:36:07to take weeks population.
00:36:12Machine they are dedicated losers that are designed for simulation.
00:36:22There is super computer and the us con Anton Gualdi.
00:36:27Hardware is designed to
00:36:30this time to speed up the process system of.
00:36:37Encoding the state of operations directly in the hardware now that
00:36:44is also a few to apply.
00:36:50Quantum computing to the state of life and works quite.
00:36:58quadri? Okay, so what you see with the simulation,
00:37:04of course on the amount of time you have to nations
00:37:11and you see what you see if for instance you
00:37:15simulate your protein one second one piece,
00:37:19second is the twelve seconds,
00:37:24you only see vibrations so you tube,
00:37:29molecular water that are you see.
00:37:36Along optimal.
00:37:46You see John Bond breaking.
00:37:51Or diffusion processes am 40 seconds dai paul effects
00:38:01dai paul effects of elettrostatiche propagate across
00:38:08the end of change for
00:38:13instance 600 you need to simulate at least then second.
00:38:20A protein movement and second second hand that you
00:38:29can see information and transitions domains or
00:38:33stalkers change or folding and local level.
00:38:39Robert Robert Folding is more than second.
00:38:45Other questions.
00:38:49Is.
00:39:00So what we
00:39:04simulate information that used to logical problems just
00:39:14added a few references is very easy way to to
00:39:21understand this is a collection of
00:39:23video YouTube from personal force.
00:39:27You can do you get the.
00:39:31Understanding of the mathematics behind
00:39:36emotions on just to you very simple example.
00:39:45Then human an create place a.
00:39:51Example. So you see one of
00:39:59di di application could be to validate model
00:40:05be you model using a template below.
00:40:13Have you also have you want to see your model
00:40:17actually keep it is your model.
00:40:21Actually is the name simulation pending on the below sites or die
00:40:30momentum mission files and also other parameters like that actually
00:40:40affect the speed so increase di di di temperature actually increase
00:40:46the momentum because this is of adesso.
00:40:56So we can actually test if then the main or not.
00:41:05So you see you have you have this protein that in
00:41:10this is to DNA molecular short e
00:41:16it is sequenze di DNA
00:41:20and you see the model is actually similar template of everything.
00:41:29Is are some of the parameters that is
00:41:32usually when you simulate the state of this system.
00:41:39So you have to define di box
00:41:43of you see the water here we have the fine di.
00:41:50In this case it has been used hand not products you have to define
00:42:00conditions and typeof hash so used to find the second the time
00:42:11in the space seems fix that you can also increase of this
00:42:16and so you see di heating and the space and different things.
00:42:26IT is. Living to the moving from the system and
00:42:34the end of the year form or writing interventions
00:42:41and the simulation the space.
00:42:46And this is typically what you want to see?
00:42:50So what is the type of representation that you should simulation is
00:42:59not super informative simple minds for
00:43:03education distance in Armstrong or in this time square.
00:43:10This is this the distance so you see just the nation
00:43:14of your information and the concepts
00:43:18compare with di information you see
00:43:23here if the current
00:43:25structures David Yates from the initial structure.
00:43:29This is the way out to
00:43:33see you think as we have the model and one case.
00:43:38We have the wild type so di original ha origine see is that
00:43:44the point A parte I think this is for the last,
00:43:52but the difference between di the wild type and
00:43:56the models to compare the two simulation,
00:44:01while they are am when the stars
00:44:05go to become different so you can just see you
00:44:09can look at the stars that you have you that
00:44:15this point and see they are
00:44:18different from the stars to this point during the simulation.
00:44:23So this is where the nation or distance along
00:44:28the story by the way the software gives you to the output system.
00:44:36The positions positions.
00:44:40If you want to do this type of
00:44:42population you have to extract snapshot from
00:44:46the story is like your sampling and just a few examples.
00:44:54If you. Have a few points.
00:45:01So this is over time is one
00:45:06of this is the book as you do something similar,
00:45:10but you and so you estimate how much
00:45:15as you along the simulation is over time.
00:45:22And gives you a model information about what are the Daily News?
00:45:31More. In this.
00:45:37Is the software di open source version of
00:45:44simulation software as a command line tools
00:45:49and other just to share this time.
00:45:53This workflow is to show that.
00:45:56We have specific command that you can execute from the command line
00:46:02and there are few steps that you have to follow to to to.
00:46:09To actually and the dynamic
00:46:14like generate topologie which is initial inputs
00:46:19modify a little bit di PDB input is
00:46:23pdb that it is to which is presentation of your PDB,
00:46:28not in the same as pdb.
00:46:31We have to find the box you have to the protein so fix aim.
00:46:39Di The water around the surface of the cool as well as to fix di.
00:46:48The strength if you have a few charts on the surface.
00:46:55We have to put some sort of some that will new Orleans
00:47:00the charges wise the force field will alter the stars to much.
00:47:07Gli input five for the convert everything to a new format,
00:47:15you download simulation and then you will analyse
00:47:19Analisi di simulation what you want to
00:47:23see you don't think about this was just to to show,
00:47:33What is the basic idea? Yes.
00:47:39Anche l'intervento.
00:47:42Su is very simple is
00:47:47di you have the story so the confirmation that is over time,
00:47:52you just take snapshot meaning.
00:47:57This is not just as a db file
00:48:01is it is a vector presentation of moments so you can
00:48:07convert that it is snapshot and meaning that
00:48:11simple and you can can confirmation with the initial confirmation.
00:48:20So at the beginning.
00:48:26At the beginning is normal to see variations because.
00:48:36Then What you see is situations like this is what you
00:48:41expect of situation like this is where you
00:48:43see fluttuazione in the difference.
00:48:48The problem of this presentation is that you don't know.
00:48:53If this information is similar to this information.
00:48:57We are just saying that is the confirmation of the same distance to
00:49:01the one touch very simple presentation is
00:49:05to visual very macroscopiche effect a proper a better way to expect
00:49:14dynamic simulation is maybe clustering
00:49:18structures so you can can per the snapshot social and
00:49:24identifying if for this is similar to this
00:49:29one or it is what is to this one or assign or visual.
00:49:35Plus. Options their many
00:49:44way to innovations one way is to see the movie of.
00:49:50Problem of the movie of the stars is that
00:49:53sometimes you don't notice,
00:49:56sometimes this is very large so is
00:49:58very difficult to actually capture what
00:50:01is the big bang for information and change.
00:50:10So next block of slides is about a tool
00:50:15can be used to the analysis of.
00:50:31The questions.
00:50:40Okay, so what we know is the move to slides.
00:50:59And actually the tool that I've used to generate
00:51:03the dataset for your final for the final project,
00:51:08of course and call to action network generator.
00:51:19So there is nothing special is not to go to contact maps,
00:51:27the only difference in the tools that we seen so
00:51:32it that is it doesn't generated content maps
00:51:38just based on businesses that is generated content maps also
00:51:42considered other parameters geometrica parameters
00:51:47between interacting dominio assets and generate type maps so you
00:51:56have contact maps that capture the type of interaction is
00:52:01that you are looking for your your staff.
00:52:07So. Content map this is
00:52:13an example see a complex to proteins,
00:52:19where you have a number of interaction that of
00:52:24course sequence number of interaction or pairs within.
00:52:32Di di decenni e rappresenta il.
00:52:38Punto di Alfa change
00:52:40interaction design for the other change
00:52:45plus we have some interaction.
00:52:48Between the to change so they are
00:52:52inter change interaction as well here we have in
00:52:56place interaction and the year
00:53:00is to the present the state of interventions not as a matrix,
00:53:04but as a graph where are nodes and
00:53:09interaction are di edge of this is actually true for.
00:53:18This is.
00:53:21For this agency list of many different am.
00:53:28Un gusto pesante, insomma.
00:53:34And this is in a Matrix so the day of this is to
00:53:45capture the type of interaction is not using
00:53:49force fields sports fields production sampling actions between
00:53:59but just playing
00:54:02geometriche constraints geometriche evaluation and for instance.
00:54:11If you are you will find the.
00:54:16Mount of paper explain what should be for
00:54:22instance di ideal angle between
00:54:26the di Di Axis connecting donor and the ADCs effect and di.
00:54:35Comparto di axis connecting dell'ONU sector a
00:54:41autobus e so you can define if the single for instance.
00:54:49Whether this is not you can what the type of
00:54:56amminoacido that are the serpent distance and say if this is not.
00:55:04We can also have another type of interaction that involve
00:55:10automatic things automatic does not some.
00:55:19Change where we are the presentation of
00:55:23the carbon indicator that there
00:55:24is here a double bond There is a single bond,
00:55:28but in reality there is this clear separation in this case for this
00:55:35is the double bond properties
00:55:39is shared gross and its adesso making.
00:55:45IT is specific chemical properties and having to
00:55:50this between two different assets that are either
00:55:55stock one of the others or they are specific angle and
00:56:02specific super position near e
00:56:07scolpite a particular stable many.
00:56:16You have cases clusters of the state of
00:56:20change that are combined together to stabilite structures.
00:56:27Things like allo Zen Atom,
00:56:32allo Zen di Atom Son di al di fuori della.
00:56:41Those that labs La one balance.
00:56:48That you have also, of course.
00:56:51Allo zen and so.
00:56:56You have pipe again an automatic in
00:57:01class action Functional group here Leasing,
00:57:07leasing e minacce a am Posted.
00:57:15By as month This is.
00:57:21Metal Coordination and of course di
00:57:28Van der Waals su year you see their some that are, of course.
00:57:35I bond, i bond van der Waals di fight the
00:57:43best are just very and their
00:57:47for their very interesting for the structure biology.
00:57:52They are not usually capture well in my field or the capture
00:58:00explicit by forced so
00:58:03the question is what we need this life of the presentation.
00:58:08Whether there is a number of applications that are
00:58:12associated to this and how it
00:58:17simply what as you do to capture sono semplici problem and this is
00:58:26very good presentation the presentation of something that is
00:58:31a continuum that because this is where
00:58:35you have things that you can can you for feeding.
00:58:39Machine learning methods Olympics so
00:58:43just a few more details
00:58:46am and this is just a più di tales about how works.
00:58:56So if you the member this is pdf format you are di type of atom.
00:59:05This is a classical atom.
00:59:08Coordinate You have the time of assets atom.
00:59:14Actually. To the line we have the Name of the carbon carbon and
00:59:24the other carbon am you have
00:59:29di aim asimmetrica it is like it, this is more.
00:59:37Precise sequenze di the position sequence and so.
00:59:49So the question is what we have coordinate,
00:59:54but in order to calcolate geometriche properties.
00:59:57We have to define what how are connected,
01:00:01of course we know that we are talking about
01:00:03using we know that the carbon oxygen
01:00:08the art specific way out how can we
01:00:13inform computer about this so this position how they are Connected.
01:00:24So di only way to construct this information.
01:00:30Gli aminoacidi aminoacidi altro sangue.
01:00:36But also if.
01:00:39You have modified as you have to fans di
01:00:44Hitler groups and so on so this addiction.
01:00:49This is the chemical component
01:00:52where you have the information about who is
01:00:57connected with you and you see you have it for
01:01:00every group and the PDP so you are we have.
01:01:11So for every compound it that we have years In
01:01:17every pdp we have in this dictionary telos information about.
01:01:27Us connected to the.
01:01:35Hotel is because this is your nitrogen connected
01:01:42to alpha carbon and then you and so on.
01:01:53The problem is that in cases.
01:01:58We know instance should be attached to to something but.
01:02:05IT. Is the solution.
01:02:11We don't have the coordinate of the business
01:02:13so another problem And
01:02:19actually we can the light that we can also code.
01:02:25We are a few endless endless distance
01:02:31because we know that it is to be
01:02:35Maxisize to be has to be place in a position that
01:02:40Maxisize the distance between all other options.
01:02:47And this is through also for every atom.
01:02:52Scoperto il nuovo modulo because we want to
01:02:57Maxisize the distance between elections that are you.
01:03:04See you see in this case.
01:03:07We have to find this twelve specific di size.
01:03:15Minerals and we specifi
01:03:20who are the other to tons that we should consider to
01:03:24this angle and other we have to used to
01:03:28the second hand with first place,
01:03:34you see you to
01:03:38this access and we have along perpendicolare axis of the.
01:03:46Axis by the same.
01:03:50Dio. Ci aiuti a Bond Bond England Bond.
01:03:57Buon. Week end What we have the absence place?
01:04:04We can can be the way to get.
01:04:11A thing is to understand.
01:04:18When we have a automatic thing and it's not to have to know who is
01:04:28connected to be we have to employment deep first search to
01:04:35to find out that the artist that from one to the thing.
01:04:45To get back to the same.
01:04:48This is it is an corrisponde
01:04:55to the response to make
01:04:58the other things that we have to that is for instance.
01:05:02If you are here and we can do product of the coordinate non sono
01:05:11so there is number of things that you have to do it is so
01:05:19the other problem of this software is also
01:05:24di calculator time so again, we have to.
01:05:30All possible outcomes because we are
01:05:33building with non covalenti interaction.
01:05:36Sapore di bonds, of course,
01:05:40and so what you could do instead of actually
01:05:44performing over you can defined
01:05:49boxes or bound boxes
01:05:54and for instance you can keep all the comparison between that are
01:05:59outside the two boxes that are interacting so
01:06:03you boxes like keeping aim
01:06:07Sangue a fiumi in assenza di coordinate di edge of this boxes and.
01:06:24So on is that things interaction network generator so actually it.
01:06:35Is the complexity of D am contacts contact map by the present in
01:06:46ways so we have seen for instance with that we can am
01:06:54contacts between the asset by
01:06:57setting keeping information about the closest players between.
01:07:05We can be more precise than that we are we have we are.
01:07:11We can increase the number of
01:07:15ages between two aminoacidi for instance.
01:07:18We see that is to
01:07:21aminoacidi can perform more than two interventions.
01:07:26Also is not actually easy to distinguished when you have one of.
01:07:35The last dance for being another types am
01:07:45Esso gives you di um option to the things in
01:07:51different ways you can the one which is the most energetic.
01:08:01Bond I can the.
01:08:04Porte multiple edges between this this is the default.
01:08:12And all edges for instance you have to where you have
01:08:18a number button below the end of the interaction between.
01:08:27To or you can be the best each type so we have then. Van der Waals.
01:08:36Introduction to the most.
01:08:43Of this for.
01:08:46Types of the presentation can be used in different context and.
01:08:56Can you different types of information that the other.
01:09:05Application is the type of atom that you want to
01:09:12consider you can consider
01:09:17just carbon carbon or what is called lollipop.
01:09:22This is just the center of the site
01:09:25so what is the closest or the other things.
01:09:31AM. Hai? Eh? Eh. Eh. Ehm. Sì. Stage for
01:09:41interaction contacts so in this case, of course.
01:09:45You can not calcolate specific types of you are just. Contact maps.
01:09:53Distance to this points distance when you consider all type of
01:10:02minacce di non generic interaction
01:10:08evaluate all options between two aminoacidi.
01:10:12Of course with this different types
01:10:16of presentation and cardinalità in autobus.
01:10:20So just some statistics on the PDB.
01:10:27So the number is the distribution different types of interaction.
01:10:37So the other thing is that you can
01:10:40distinguished between aminoacidi interaction and
01:10:43aminoacidi with interaction and
01:10:49you can also distinguished in between interface interaction.
01:10:54M.B. What's interesting here is
01:10:56the different Situations presence I bonds.
01:11:05They are common in
01:11:09intra chain interaction between common assets and also legal
01:11:17legal in legal modified assets
01:11:23or see also the number of the number of contacts
01:11:28is very law here we are mine millions are we
01:11:31have that thousand aim in
01:11:36the TAR Prodi more interesting to see the interfaces you see
01:11:44Senti six percent and
01:11:4820% evidence and also you can see them nucleare assets.
01:11:54Il nostro nucleo
01:11:55Asset Actions Aminoacidi Nuclei Asset interesse nei confronti
01:12:02del bonus bebè che si intreccia in interaction and so meaning that
01:12:12automatic link with you to connect to
01:12:18different surface is maybe also pick and pay,
01:12:24but also ionic bond more.
01:12:28Interaction is so what we can get from
01:12:33the table is that it is more use for the.
01:12:40Other type of interaction or used or surface is
01:12:45to make the times for the form complex.
01:12:52Okay, so how we can apply this time to
01:12:57Nations we can actually employment what are
01:13:01called probabilistiche based interaction network
01:13:05and the idea is very simple
01:13:08you create graph for every model structure probability.
01:13:16Probability interaction along the various model or Snapchat.
01:13:24Simulation. This is what you can also find
01:13:33in this is an output of
01:13:38the way you have the command line is C++ software.
01:13:46Leader Compile in download Compile version.
01:13:51Plugin. Final Cut and
01:14:01his server actually see the interaction on
01:14:05the fly on the server so we can actually
01:14:10can you an example I don't have
01:14:13simulation test and we have a more models.
01:14:23So di interfaces various parameters
01:14:30the street art and manual parameters that are the light
01:14:35to the geometrica use to the contacts,
01:14:41you can also and other things like
01:14:44whether want to water molecular or.
01:14:48Ad or don't sottomultipli
01:14:53Di output file di files file su file list
01:15:02notes that where process during
01:15:04the temptation edges are di
01:15:08the list of file. Presentation.
01:15:26Statistics about how parameters where identifier
01:15:32so you simply distance is an observer what are the pigs?
01:15:37We can calcolate al di sotto?
01:15:47Di distribution and.
01:15:51Dice A time of the member States of time many to see
01:16:00weather some of the this is actually
01:16:02the two second for specific bird.
01:16:09Yes, we can up with this is after check in examples and so you know
01:16:19how to give you don't have you have
01:16:24just a capo of slides that are
01:16:26detail about how to the server employment,
01:16:30which think of interest for you.
01:16:38And applications all the web server it's
01:16:43dedicate in modo I want to do not.
01:16:50On the website should be.
01:17:00It's. Dates probably one problem
01:17:10because this definition are switch off the server.
01:17:16So.
01:17:27Che. This is running.
01:17:40Okay, sorry as they are.
01:17:46Their using the.
01:17:49Update in the server the decided to do it
01:17:52today so there is no way can you.
01:17:57Could show the plugin but the plugin actually uses the server
01:18:04to to the contacts at the think thank you,
01:18:12then is d assignment
01:18:16for the final project because introduce almost thing.
01:18:21I want to to introduce e can start with the way, it is.
01:18:54Also make it variable.
01:19:21You should be access to this file.
01:19:37Ok, so am abbiamo size tomorrow you more About it could.
01:19:48So the idea of the project is to be able to provide
01:19:57type interaction starting from di output of a.
01:20:06Python so you can calcolate a contact map from
01:20:11by Python just based on this and the question
01:20:16is can you design
01:20:19employment device model machine learning model that can
01:20:23classified contacts automatic based
01:20:28on features that you can extract from the structures.
01:20:34So what I did it extract some of
01:20:38the features for you or at least provide some code
01:20:43to extract features provided
01:20:47training dataset so what you ask to do is to
01:20:53design a. Multilevel classification model or one versus or
01:21:03whatever you want to employment to give a probability given
01:21:11a contact between the two assets
01:21:16of being one of this type of interaction.
01:21:24The year is you have to be careful
01:21:29because that the same per this is not the first here.
01:21:34I have this is not.
01:21:39The same of mind asset che make more than one type of
01:21:44interaction to be careful with that so.
01:21:57Have software which can be used to extract the features.
01:22:04I have with the training,
01:22:06which is quite large have a lot of examples.
01:22:11Also an example of am a simple classified very
01:22:19have bias classified to show
01:22:23that you you should do better than that.
01:22:27And that's it. There is not that I will explain tomorrow,
01:22:33which is the three di Alphabet,
01:22:39which is something that is extract through a transformer model and
01:22:46its provided year body transformer model and
01:22:50di conference a piece of software that can extract that alphabet,
01:22:56but to more what is the the meaning?
01:22:59Why it is it can be very useful.
01:23:03And. ITS so about
01:23:09the features of the training set you see the amount of
01:23:14a type of contact that you have you have
01:23:17one about one billion bonds seven and so this would be
01:23:23the most complicate most of them overlap
01:23:28to be careful because if you build
01:23:31a single model that products just one class.
01:23:34You of everything is one of the two classes,
01:23:38Special if you implementa soft max and you
01:23:42get the maximum probability just for one of them and
01:23:46also you have a number of things that
01:23:50are an classified what I mean so how I build
01:23:55a dataset I calculator contact maps using Python so measure
01:24:01ai found Peirce used that have at least one atom below.
01:24:11And then assigned di ring classes
01:24:16was able to classified one million
01:24:21of this contact IT doesn't mean that you have to
01:24:26employment a classified that is
01:24:31able to classified classified contacts may be those are just
01:24:37Positive thing Maybe you have
01:24:40a nitrogen bond that is a faction of Armstrong,
01:24:44just about the old gets classified by by Renzo I could happen.
01:24:52Or they can also be.
01:24:56Non force contact or types of
01:25:00contact or not capture by simply not capture by admin
01:25:05So careful you have to take decisions about the way you you
01:25:10analyse the dataset so the type of features that could be useful.
01:25:18So you have so a di as underscore or the T underscore
01:25:28is for source and target simply indicate
01:25:33if Are you have
01:25:37one is the source and the other is target, of course it.
01:25:40IT can be also switch and in the training
01:25:44set di duplicate switching the order.
01:25:51So you have the identification of of di aminoacidi so
01:25:58chain name index it is located in di in di in
01:26:06di in the sequence insertion con residuo name again index.
01:26:15AM it's tricky to use you
01:26:20should too much just using
01:26:23the index because the only things could be used.
01:26:26Maybe is the difference between the index.
01:26:28Just do check to contacts are close or distant
01:26:32in the in the in the sequence plus
01:26:36team di other features are secondary structure
01:26:39eight states provided from DSP.
01:26:46Accessibility files di features
01:26:53The member features capture properties
01:26:57asset like to the city and other things and three di Alphabet
01:27:07the state and the alphabet at the same things so you
01:27:10use one is numerica and the other is
01:27:14alphabet that you have the same for the second the second menace
01:27:23so you have the software this features
01:27:27meaning they are ricalcolate for your training that set.
01:27:34Also the software to do you are you should integrate also
01:27:40di feature extraction in your influence piece of
01:27:45software and your free to the other features so if you want
01:27:50to be different things even running by Python,
01:27:56whatever you want to go to extract.
01:27:58Other properties that have been over look by me in this exercise.
01:28:03You are welcome to do you.
01:28:07So what you request to do to provide.
01:28:12As a said it should give given
01:28:16an interaction capture by by
01:28:19Python based on distance you should tell me,
01:28:23what is the type of that interaction and a probability sign.
01:28:32The model description and
01:28:36a report that could explain what you did so short report?
01:28:41It's like a paper just a few pages,
01:28:44you see between two five pages of Tex,
01:28:47please the right to draft it as it.
01:28:50IT was a paper and you will discuss you can provide also
01:28:57some statistics some plot and some analysis as well
01:29:01as about what's your training strategy?
01:29:06What is the algorithm that you tested.
01:29:10If they were something that is better than other things and so on
01:29:15and documentation so can understand that
01:29:20you want you provided under the use and execute it.
01:29:26A property so well evaluate the performance of
01:29:30the software that so how difficult is
01:29:34for me to your software and
01:29:38the documentation so about this is the nation of di.
01:29:44Dialogo di usage software as well as
01:29:49di a good is the department of you under the problem.
01:29:55Ok stop it here to know if you have questions.
01:30:02Should be web server fully and also you di Transformers
01:30:11things that have been applied to play and
01:30:16also di generation of
01:30:17the three alphabet and what is the meaning of the alphabet.
01:30:22Ok Tomorrow.